The Definitive Guide, 2nd Edition

The Definitive Guide, 2nd Edition
802.11® Wireless Networks The Definitive Guide
By Matthew Gast
...............................................
Publisher: O'Reilly
Pub Date: April 2005
ISBN: 0-596-10052-3
Pages: 656
Table of Contents | Index
As we all know by now, wireless networks offer many advantages over fixed (or wired) networks.
Foremost on that list is mobility, since going wireless frees you from the tether of an Ethernet cable
at a desk. But that's just the tip of the cable-free iceberg. Wireless networks are also more flexible,
faster and easier for you to use, and more affordable to deploy and maintain.
The de facto standard for wireless networking is the 802.11 protocol, which includes Wi-Fi (the
wireless standard known as 802.11b) and its faster cousin, 802.11g. With easy-to-install 802.11
network hardware available everywhere you turn, the choice seems simple, and many people dive
into wireless computing with less thought and planning than they'd give to a wired network. But it's
wise to be familiar with both the capabilities and risks associated with the 802.11 protocols. And
802.11 Wireless Networks: The Definitive Guide, 2nd Edition is the perfect place to start.
This updated edition covers everything you'll ever need to know about wireless technology.
Designed with the system administrator or serious home user in mind, it's a no-nonsense guide for
setting up 802.11 on Windows and Linux. Among the wide range of topics covered are discussions
on:
deployment considerations
network monitoring and performance tuning
wireless security issues
how to use and select access points
network monitoring essentials
wireless card configuration
security issues unique to wireless networks
With wireless technology, the advantages to its users are indeed plentiful. Companies no longer
have to deal with the hassle and expense of wiring buildings, and households with several
computers can avoid fights over who's online. And now, with 802.11 Wireless Networks: The
Definitive Guide, 2nd Edition, you can integrate wireless technology into your current infrastructure
with the utmost confidence.
802.11® Wireless Networks The Definitive Guide
By Matthew Gast
...............................................
Publisher: O'Reilly
Pub Date: April 2005
ISBN: 0-596-10052-3
Pages: 656
Table of Contents | Index
Copyright
Foreword
Preface
Prometheus Untethered: The Possibilities of Wireless LANs
Audience
Overture for Book in Black and White, Opus 2
Conventions Used in This Book
How to Contact Us
Safari Enabled
Acknowledgments
Chapter 1. Introduction to Wireless Networking
Why Wireless?
What Makes Wireless Networks Different
A Network by Any Other Name...
Chapter 2. Overview of 802.11 Networks
IEEE 802 Network Technology Family Tree
802.11 Nomenclature and Design
802.11 Network Operations
Mobility Support
Chapter 3. 802.11 MAC Fundamentals
Challenges for the MAC
MAC Access Modes and Timing
Contention-Based Access Using the DCF
Fragmentation and Reassembly
Frame Format
Encapsulation of Higher-Layer Protocols Within 802.11
Contention-Based Data Service
Frame Processing and Bridging
Chapter 4. 802.11 Framing in Detail
Data Frames
Control Frames
Management Frames
Frame Transmission and Association and Authentication States
Chapter 5. Wired Equivalent Privacy (WEP)
Cryptographic Background to WEP
WEP Cryptographic Operations
Problems with WEP
Dynamic WEP
Chapter 6. User Authentication with 802.1X
The Extensible Authentication Protocol
EAP Methods
802.1X: Network Port Authentication
802.1X on Wireless LANs
Chapter 7. 802.11i: Robust Security Networks, TKIP, and CCMP
The Temporal Key Integrity Protocol (TKIP)
Counter Mode with CBC-MAC (CCMP)
Robust Security Network (RSN) Operations
Chapter 8. Management Operations
Management Architecture
Scanning
Authentication
Preauthentication
Association
Power Conservation
Timer Synchronization
Spectrum Management
Chapter 9. Contention-Free Service with the PCF
Contention-Free Access Using the PCF
Detailed PCF Framing
Power Management and the PCF
Chapter 10. Physical Layer Overview
Physical-Layer Architecture
The Radio Link
RF Propagation with 802.11
RF Engineering for 802.11
Chapter 11. The Frequency-Hopping (FH) PHY
Frequency-Hopping Transmission
Gaussian Frequency Shift Keying (GFSK)
FH PHY Convergence Procedure (PLCP)
Frequency-Hopping PMD Sublayer
Characteristics of the FH PHY
Chapter 12. The Direct Sequence PHYs: DSSS and HR/DSSS (802.11b)
Direct Sequence Transmission
Differential Phase Shift Keying (DPSK)
The "Original" Direct Sequence PHY
Complementary Code Keying
High Rate Direct Sequence PHY
Chapter 13. 802.11a and 802.11j: 5-GHz OFDM PHY
Orthogonal Frequency Division Multiplexing (OFDM)
OFDM as Applied by 802.11a
OFDM PLCP
OFDM PMD
Characteristics of the OFDM PHY
Chapter 14. 802.11g: The Extended-Rate PHY (ERP)
802.11g Components
ERP Physical Layer Convergence (PLCP)
ERP Physical Medium Dependent (PMD) Layer
Chapter 15. A Peek Ahead at 802.11n: MIMO-OFDM
Common Features
WWiSE
TGnSync
Comparison and Conclusions
Chapter 16. 802.11 Hardware
General Structure of an 802.11 Interface
Implementation-Specific Behavior
Reading the Specification Sheet
Chapter 17. Using 802.11 on Windows
Windows XP
Windows 2000
Windows Computer Authentication
Chapter 18. 802.11 on the Macintosh
The AirPort Extreme Card
802.1X on the AirPort
Chapter 19. Using 802.11 on Linux
PCMCIA Support on Linux
Linux Wireless Extensions and Tools
Agere (Lucent) Orinoco
Atheros-Based cards and MADwifi
802.1X on Linux with xsupplicant
Chapter 20. Using 802.11 Access Points
General Functions of an Access Point
Power over Ethernet (PoE)
Selecting Access Points
Cisco 1200 Access Point
Apple AirPort
Chapter 21. Logical Wireless Network Architecture
Evaluating a Logical Architecture
Topology Examples
Choosing Your Logical Architecture
Chapter 22. Security Architecture
Security Definition and Analysis
Authentication and Access Control
Ensuring Secrecy Through Encryption
Selecting Security Protocols
Rogue Access Points
Chapter 23. Site Planning and Project Management
Project Planning and Requirements
Network Requirements
Physical Layer Selection and Design
Planning Access-Point Placement
Using Antennas to Tailor Coverage
Chapter 24. 802.11 Network Analysis
Network Analyzers
Ethereal
802.11 Network Analysis Checklist
Other Tools
Chapter 25. 802.11 Performance Tuning
802.11 Performance Calculations
Improving Performance
Tunable 802.11 Parameters
Chapter 26. Conclusions and Predictions
Standards Work
Current Trends in Wireless Networking
The End
glossary
About the Author
Colophon
Index
802.11® Wireless Networks: The Definitive Guide, Second Edition
by Matthew S. Gast
Copyright © 2005 Matthew S. Gast. All rights reserved.
Printed in the United States of America.
Published by O'Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472.
O'Reilly books may be purchased for educational, business, or sales promotional use. Online editions
are also available for most titles (safari.oreilly.com). For more information, contact our
corporate/institutional sales department: (800) 998-9938 or corporate@oreilly.com.
Editor:
Mike Loukides
Production Editor:
Colleen Gorman
Cover Designer:
Ellie Volckhausen
Interior Designer:
David Futato
Printing History:
April 2002:
First Edition.
April 2005:
Second Edition.
Nutshell Handbook, the Nutshell Handbook logo, and the O'Reilly logo are registered trademarks of
O'Reilly Media, Inc. 802.11® Wireless Networks: The Definitive Guide, Second Edition, the image of a
horseshoe bat, and related trade dress are trademarks of O'Reilly Media, Inc.
802.11® and all 802.11-based trademarks and logos are trademarks or registered trademarks of
IEEE, Inc. in the United States and other countries. O'Reilly Media, Inc. is independent of IEEE.
Many of the designations used by manufacturers and sellers to distinguish their products are claimed
as trademarks. Where those designations appear in this book, and O'Reilly Media, Inc. was aware of
a trademark claim, the designations have been printed in caps or initial caps.
While every precaution has been taken in the preparation of this book, the publisher and author
assume no responsibility for errors or omissions, or for damages resulting from the use of the
information contained herein.
ISBN: 0-596-10052-3
[M]
Foreword
Matthew Gast was my mentor long before I met him. I began reporting on wireless data networking
in October 2000 when I discovered that Apple's claims for its 802.11b-based AirPort Base Station
were actually true.
I'd been burned with another form of wireless networking that used infrared, and had spent many
fruitless hours using other "interesting" networking technologies that led to dead ends. I figured
802.11b was just another one. Was I glad I was wrong!
This discovery took me down a path that led, inexorably, to the first edition of 802.11 Wireless
Networks. How did this stuff actually work as advertised? I knew plenty about the ISO model, TCP/IP,
and Ethernet frames, but I couldn't reconcile a medium in which all parties talked in the same space
with what I knew about Ethernet's methods of coping with shared contention.
Matthew taught me through words and figures that I didn't originally understand, but returned to
again and again as I descended further into technical detail in my attempts to explain Wi-Fi to a
broader and broader audience through articles in The New York Times, The Seattle Times, PC World,
and my own Wi-Fi Networking News (http://www.wifinetnews.com) site over the last five years.
I starting learning acronyms from 802.11 Wireless Networks and used Matthew's book to go beyond
expanding WDS into Wireless Distribution System into understanding precisely how two access points
could exchange data with each other through a built-in 802.11 mechanism that allowed four parties
to a packet's transit.
Now as time went by and the 802.11 family grew and became baroque, the first edition of this title
started feeling a little out of datealthough it remained surprising how many "new" innovations were
firmly rooted in developments of the early to mid-1990s. The alphabet soup of the first edition was
gruel compared to the mulligatawny of 2005.
Matthew filled the gap between the book and contemporary wireless reality through his ongoing
writing at O'Reilly's Wireless DevCenter, which I read avidly. And somewhere in there I was
introduced to Matthew at a Wi-Fi Planet conference. We hit it off immediately: I started pestering him
for details about 802.1X, if I remember correctly, and he wanted to talk about books and business. (I
wound up writing two editions of a general market Wi-Fi book, neither of which did nearly as well as
Matthew's extraordinarily technical one.)
Since then, I have been in the rare and privileged position to be the recipient of Matthew's generosity
with his knowledge and humble insight. Matthew isn't one who assumes; he researches. His natural
curiosity compels him to dig until he gets an answer that's technically and logically consistent.
Take, for instance, the incredibly political and complicated evolution of the 802.1X standard. (I know,
from Matthew, that it's properly capitalized since it's a freestanding standard not reliant on other
specifications. Even the IEEE makes this mistake, and it's their rule for capitalization that we're both
following.)
802.1X is simple enough in its use of the Extensible Authentication Protocol, a generic method of
passing messages among parties to authentication. But the ways in which EAP is secured are, quite
frankly, insanereflecting Microsoft and Cisco's parallel but conflicting attempts to control support of
legacy protocols in a way that only damages easy access to its higher level of security.
Matthew eschewed the religious debate and spelled out the various methods, difficulties, and
interoperability issues in an O'Reilly Network article that's the nugget of the expanded coverage in
this book. I defy any reader to find as cogent and exhaustive an explanation before this book was
published. There's nothing as clear, comprehensive, and unaffected by market politics.
At times, Matthew bemoaned the delays that led to the gap between editions of this book, due partly
to his joining a startup wireless LAN switch company, but I think readers are better served through
his very hard-won, late-night, long-hours knowledge.
Matthew's relationship with 802.11 might have previously been considered that of a handy man who
knew his way around the infrastructure of his house. If a toilet was running, he could replace a valve.
If the living room needed new outlets, he could research the process and wire them in.
But Matthew's new job took him allegorically from a weekend household warrior to a jack-of-alltradesman. Matthew can tear out those inner walls, reframe, plumb, and wire them, all the while
bitching about the local building code.
It's been a pleasure knowing Matthew, and it's even more a pleasure to introduce you to his book,
and let you all in on what I and others have been more private recipients of for the last few years.
Glenn Fleishman
Seattle, Washington
February 2005
Preface
People move. Networks don't.
More than anything else, these two statements can explain the explosion of wireless LAN hardware.
In just a few years, wireless LANs have grown from a high-priced, alpha-geek curiosity to
mainstream technology.
By removing the network port from the equation, wireless networks separate user connectivity from
a direct physical location at the end of a cord. To abstract the user location from the network,
however, requires a great deal of protocol engineering. For users to have location-independent
services, the network must become much more aware of their location.
This book has been written on more airplanes, in more airports, and on more trains than I care to
count. Much of the research involved in distilling evolving network technology into a book depends on
Internet access. It is safe to say that without ubiquitous network access, the arrival of this book
would have been much delayed.
The advantages of wireless networks has made them a fast-growing multibillion dollar equipment
market. Wireless LANs are now a fixture on the networking landscape, which means you need to
learn to deal with them.
Prometheus Untethered: The Possibilities of Wireless
LANs
Wireless networks offer several advantages over fixed (or "wired") networks:
Mobility
Users move, but data is usually stored centrally, enabling users to access data while they are
in motion can lead to large productivity gains. Networks are built because they offer valuable
services to users. In the past, network designers have focused on working with network ports
because that is what typically maps to a user. With wireless, there are no ports, and the
network can be designed around user identity.
Ease and speed of deployment
Many areas are difficult to wire for traditional wired LANs. Older buildings are often a problem;
running cable through the walls of an older stone building to which the blueprints have been
lost can be a challenge. In many places, historic preservation laws make it difficult to carry out
new LAN installations in older buildings. Even in modern facilities, contracting for cable
installation can be expensive and time-consuming.
Flexibility
No cables means no recabling. Wireless networks allow users to quickly form amorphous, small
group networks for a meeting, and wireless networking makes moving between cubicles and
offices a snap. Expansion with wireless networks is easy because the network medium is
already everywhere. There are no cables to pull, connect, or trip over. Flexibility is the big
selling point for the "hot spot" market, composed mainly of hotels, airports, train stations (and
even trains themselves!), libraries, and cafes.
Cost
In some cases, costs can be reduced by using wireless technology. As an example, 802.11®
equipment can be used to create a wireless bridge between two buildings. Setting up a wireless
bridge requires some initial capital cost in terms of outdoor equipment, access points, and
wireless interfaces. After the initial capital expenditure, however, an 802.11-based, line-ofsight network will have only a negligible recurring monthly operating cost. Over time, point-topoint wireless links are far cheaper than leasing capacity from the telephone company.
Until the completion of the 802.11 standard in 1997, however, users wanting to take advantage of
these attributes were forced to adopt single-vendor solutions with all of the risk that entailed. Once
802.11 started the ball rolling, speeds quickly increased from 2 Mbps to 11 Mbps to 54 Mbps.
Standardized wireless interfaces and antennas have made it much easier to build wireless networks.
Several service providers have jumped at the idea, and enthusiastic bands of volunteers in most
major cities have started to build public wireless networks based on 802.11.
802.11 has become something of a universally assumed connectivity method as well. Rather than
wiring public access ports up with Ethernet, a collection of access points can provide connectivity to
guests. In the years since 802.11 was standardized, so-called "hot spots" have gone from an exotic
curiosity in venues that do not move, to technology that is providing connectivity even while in
transit. By coupling 802.11 access with a satellite uplink, it is possible to provide Internet access
even while moving quickly. Several commuter rail systems provide mobile hot-spots, and Boeing's
Connexion service can do the same for an airplane, even at a cruising speed of 550 miles per hour.
Audience
This book is intended for readers who need to learn more about the technical aspects of wireless
LANs, from operations to deployment to monitoring:
Network architects contemplating rolling out 802.11 equipment onto networks or building
networks based on 802.11
Network administrators responsible for building and maintaining 802.11 networks
Security professionals concerned about the exposure from deployment of 802.11 equipment
and interested in measures to reduce the security headaches
The book assumes that you have a solid background in computer networks. You should have a basic
understanding of IEEE 802 networks (particularly Ethernet), the OSI reference model, and the TCP/IP
protocols, in addition to any other protocols on your network. Wireless LANs are not totally new
ground for most network administrators, but there will be new concepts, particularly involving radio
transmissions.
Overture for Book in Black and White, Opus 2
Part of the difficulty in writing a book on a technology that is evolving quickly is that you are never
quite sure what to include. The years between the first and second edition were filled with many
developments in security, and updating the security-related information was one of the major parts
of this revision. This book has two main purposes: it is meant to teach the reader about the 802.11
standard itself, and it offers practical advice on building wireless LANs with 802.11 equipment. These
two purposes are meant to be independent of each other so you can easily find what interests you.
To help you decide what to read first and to give you a better idea of the layout, the following are
brief summaries of all the chapters.
Chapter 1, Introduction to Wireless Networking, lists ways in which wireless networks are different
from traditional wired networks and discusses the challenges faced when adapting to fuzzy
boundaries and unreliable media. Wireless LANs are perhaps the most interesting illustration of
Christian Huitema's assertion that the Internet has no center, just an ever-expanding edge. With
wireless LAN technology becoming commonplace, that edge is now blurring.
Chapter 2, Overview of 802.11 Networks, describes the overall architecture of 802.11 wireless LANs.
802.11 is somewhat like Ethernet but with a number of new network components and a lot of new
acronyms. This chapter introduces you to the network components that you'll work with. Broadly
speaking, these components are stations (mobile devices with wireless cards), access points
(glorified bridges between the stations and the distribution system), and the distribution system itself
(the wired backbone network). Stations are grouped logically into Basic Service Sets (BSSs). When
no access point is present, the network is a loose, ad-hoc confederation called an independent BSS
(IBSS). Access points allow more structure by connecting disparate physical BSSs into a further
logical grouping called an Extended Service Set (ESS).
Chapter 3, 802.11 MAC Fundamentals, describes the Media Access Control (MAC) layer of the 802.11
standard in detail. 802.11, like all IEEE 802 networks, splits the MAC-layer functionality from the
physical medium access. Several physical layers exist for 802.11, but the MAC is the same across all
of them. The main mode for accessing the network medium is a traditional contention-based access
method, though it employs collision avoidance (CSMA/CA) rather than collision detection (CSMA/CD).
The chapter also discusses data encapsulation in 802.11 frames and helps network administrators
understand the frame sequences used to transfer data.
Chapter 4, 802.11 Framing in Detail, builds on the end of Chapter 3 by describing the various frame
types and where they are used. This chapter is intended more as a reference than actual reading
material. It describes the three major frame classes. Data frames are the workhorse of 802.11.
Control frames serve supervisory purposes. Management frames assist in performing the extended
operations of the 802.11 MAC. Beacons announce the existence of an 802.11 network, assist in the
association process, and are used for authenticating stations.
Chapter 5, Wired Equivalent Privacy (WEP), describes the Wired Equivalent Privacy protocol. In spite
of its flaws, WEP is the basis for much of the following work in wireless LAN security. This chapter
discusses what WEP is, how it works, and why you can't rely on it for any meaningful privacy or
security.
Chapter 6, User Authentication with 802.1X, describes the 802.1X authentication framework. In
conjunction with the Extensible Authentication Protocol, 802.1X provides strong authentication
solutions and improved encryption on Wireless LANs.
Chapter 7, 802.11i: Robust Security Networks, TKIP, and CCMP, describes the 802.11i standard for
wireless LAN security. In recognition of the fundamental flaws of WEP, two new link-layer encryption
protocols were designed, complete with new mechanisms to derive and distribute keys.
Chapter 8, Management Operations, describes the management operations on 802.11 networks. To
find networks to join, stations scan for active networks announced by access points or the IBSS
creator. Before sending data, stations must associate with an access point. This chapter also
discusses the power-management features incorporated into the MAC that allow battery-powered
stations to sleep and pick up buffered traffic at periodic intervals.
Chapter 9, Contention-Free Service with the PCF, describes the point coordination function. The PCF
is not widely implemented, so this chapter can be skipped for most purposes. The PCF is the basis for
contention-free access to the wireless medium. Contention-free access is like a centrally controlled,
token-based medium, where access points provide the "token" function.
Chapter 10, Physical Layer Overview, describes the general architecture of the physical layer (PHY) in
the 802.11 model. The PHY itself is broken down into two "sublayers." The Physical Layer
Convergence Procedure (PLCP) adds a preamble to form the complete frame and its own header,
while the Physical Medium Dependent (PMD) sublayer includes modulation details. The most common
PHYs use radio frequency (RF) as the wireless medium, so the chapter closes with a short discussion
on RF systems and technology that can be applied to any PHY discussed in the book.
Chapter 11, The Frequency-Hopping (FH) PHY, describes the oldest physical layer with 802.11.
Products based on the FH PHY are no longer widely sold, but a great deal of early 802.11 equipment
was based on them. Organizations with a long history of involvement with 802.11 technology may
need to be familiar with this PHY.
Chapter 12, The Direct Sequence PHYs: DSSS and HR/DSSS (802.11b), describes two physical layers
based on direct sequence spread spectrum technology. The initial 802.11 standard included a layer
which offered speeds of 1 Mbps and 2 Mbps. While interesting, it was not until 802.11b added 5.5
Mbps and 11 Mbps data rates that the technology really took off. This chapter describes the two
closely-related PHYs as a single package.
Chapter 13, 802.11a and 802.11j: 5-GHz OFDM PHY, describes the 5-GHz PHY standardized with
802.11a, which operates at 54 Mbps. This physical layer uses another modulation technique known
as orthogonal frequency division multiplexing (OFDM). Slight modifications were required to use this
PHY in Japan, which were made by the 802.11j standard.
Chapter 14, 802.11g: The Extended-Rate PHY (ERP), describes a PHY which uses OFDM technology,
but in the 2.4 GHz frequency band shared by 802.11b. It has largely supplanted 802.11b, and is a
common option for built-in connectivity with new notebook computers. The PHY itself is almost
identical to the 802.11a PHY. The differences are in allowing for backwards compatibility with older
equipment sharing the same frequency band.
Chapter 15, A Peek Ahead at 802.11n: MIMO-OFDM, describes the PHY currently in development.
802.11n uses a PHY based on multiple-input/multiple-output (MIMO) technology for much higher
speed. At the time this book went to press, two proposed standards were dueling in the committee.
This chapter describes both.
Chapter 16, 802.11 Hardware, begins the transition from theoretical matters based on the standards
to how the standards are implemented. 802.11 is a relatively loose standard, and allows a large
number of implementation choices. Cards may differ in their specified performance, or in the manner
in which certain protocols are implemented. Many of these variations are based on how they are
built.
Chapter 17, Using 802.11 on Windows, describes the basic driver installation procedure in Windows,
and how to configure security settings.
Chapter 18, 802.11 on the Macintosh, describes how to use the AirPort card on MacOS X to connect
to 802.11 networks. It focuses on Mac OS X 10.3, which was the first software version to include
802.1X support.
Chapter 19, Using 802.11 on Linux, discusses how to install 802.11 support on a Linux system. After
discussing how to add PC Card support to the operating system, it shows how to use the wireless
extensions API. It discusses two common drivers, one for the older Orinoco 802.11b card, and the
MADwifi driver for newer cards based on chipsets from Atheros Communications. Finally, it shows
how to configure 802.1X security using xsupplicant.
Chapter 20, Using 802.11 Access Points, describes the equipment used on the infrastructure end of
802.11 networks. Commercial access point products have varying features. This chapter describes
the common features of access points, offers buying advice, and presents two practical configuration
examples.
Chapter 21, Logical Wireless Network Architecture, marks the third transition in the book, from the
implementation of 802.11 on the scale of an individual device, to how to build 802.11 networks on a
larger scale. There are several major styles that can be used to build the network, each with its
advantages and disadvantages. This chapter sorts through the common types of network topologies
and offers advice on selecting one.
Chapter 22, Security Architecture, should be read in tandem with the previous chapter. Maintaining
network security while offering network access on an open medium is a major challenge. Security
choices and architecture choices are mutually influential. This chapter addresses the major choices to
be made in designing a network: what type of authentication will be used and how it integrates with
existing user databases, how to encrypt traffic to keep it safe, and how to deal with unauthorized
access point deployment.
Chapter 23, Site Planning and Project Management, is the final component of the book for network
administrators. Designing a large-scale wireless network is difficult because there is great user
demand for access. Ensuring that the network has sufficient capacity to satisfy user demands in all
the locations where it will be used requires some planning. Choosing locations for access points
depends a great deal on the radio environment, and has traditionally been one of the most timeconsuming tasks in building a network.
Chapter 24, 802.11 Network Analysis, teaches administrators how to recognize what's going on with
their wireless LANs. Network analyzers have proven their worth time and time again on wired
networks. Wireless network analyzers are just as valuable a tool for 802.11 networks. This chapter
discusses how to use wireless network analyzers and what certain symptoms may indicate. It also
describes how to build an analyzer using Ethereal, and what to look for to troubleshoot common
problems.
Chapter 25, 802.11 Performance Tuning, describes how network administrators can increase
throughput. It begins by describing how to calculate overall throughput for payload data, and
common ways of increasing performance. In rare cases, it may make sense to change commonly
exposed 802.11 parameters.
Chapter 26, Conclusions and Predictions, summarizes current standards work in the 802.11 working
group. After summarizing the work in progress, I get to prognosticate and hope that I don't have to
revise this too extensively in future editions.
Major Changes from the First Edition
The three years between 2002 and 2005 saw a great deal of change in wireless LANs. The standards
themselves continued to evolve to provide greater security and interoperability. Following the typical
technology path of "faster, better, and cheaper," the data rate of most 802.11 interfaces has shot
from 2 or 11 Mbps with 802.11b to 54 Mbps with 802.11a and 802.11g. Increased speed with
backwards compatibility has proved to be a commercially successful formula for 802.11g, even if it
has limitations when used for large-scale networks. The coming standardization of 802.11n is set to
boost speeds even farther. New developments in PHY technology are anxiously awaited by users, as
shown by the popular releases of pre-standard technology. Two entirely new chapters are devoted to
802.11g and 802.11n. European adoption of 802.11a was contingent on the development of
spectrum management in 802.11h, which resulted in extensive revisions to the management
chapter.
When the first edition was released in 2002, the perception of insecurity dominated discussions of the
technology. WEP was clearly insufficient, but there was no good alternative. Most network
administrators were making do with remote access systems turned inward, rather than their natural
outward orientation. The development of 802.11i was done a great deal to simplify network security.
Security is now built in to the specification, rather than something which must be added on after
getting the network right. Security improvements permeate the book, from new chapters showing
how the new protocols work, to showing how they can be used on the client side, to how to sort
through different options when building a network. Sorting through security options is much more
complex now than it was three years ago, and made it necessary to expand a section of the
deployment discussion in the first edition into its own chapter.
Three years ago, most access points were expensive devices that did not work well in large numbers.
Network deployment was often an exercise in working around the limitations of the devices of the
time. Three years later, vastly more capable devices allow much more flexible deployment models.
Rather than just a "one size fits all" deployment model, there are now multiple options to sort
through. Security protocols have improved enough that discussions of deploying technology are
based on what it can do for the organization, not on fear and how to keep it controlled. As a result,
the original chapter on network deployment has grown into three, each tackling a major part of the
deployment process.
Conventions Used in This Book
Italic is used for:
Pathnames, filenames, class names, and directories
New terms where they are defined
Internet addresses, such as domain names and URLs
Bold is used for:
GUI components
Constant Width is used for:
Command lines and options that should be typed verbatim on the screen
All code listings
Constant Width Italic is used for:
General placeholders that indicate that an item should be replaced by some actual value in your
own program
Constant Width Bold is used for:
Text that is typed in code examples by the user
Indicates a tip, suggestion, or general note
Indicates a warning or caution
How to Contact Us
Please address comments and questions concerning this book to the publisher:
O'Reilly Media, Inc.
1005 Gravenstein Highway North
Sebastopol, CA 95472
(800) 998-9938 (in the U.S. or Canada)
(707) 829-0515 (international/local)
(707) 829-0104 (fax)
There is a web site for the book, where errata and any additional information will be listed. You can
access this page at:
http://www.oreilly.com/catalog/802dot112/
In a fast-moving field, smaller articles bridge the gap between contemporary practice and the last
version of the printed book. You can access my weblog and articles at:
http://weblogs.oreillynet.com/pub/au/692/
To comment or ask technical questions about this book, send email to:
bookquestions@oreilly.com
For more information about our books, conferences, software, Resource Centers, and the O'Reilly
Network, see our web site at:
http://www.oreilly.com/
Safari Enabled
When you see a Safari® Enabled icon on the cover of your favorite technology book, it
means the book is available online through the O'Reilly Network Safari Bookshelf.
Safari offers a solution that's better than e-books. It's a virtual library that lets you easily search
thousands of top technology books, cut and paste code samples, download chapters, and find quick
answers when you need the most accurate, current information. Try it for free at
http://safari.oreilly.com.
Acknowledgments
As much as I would like to believe that you are reading this book for its entertainment value, I know
better. Technical books are valued because they get the details right, and convey them in an easier
fashion than the unadorned technical specification. Behind every technical book, there is a review
team that saw the first draft and helped to improve it. My review team caught numerous mistakes
and made the book significantly better. Dr. Malik Audeh of Tropos Networks is, for lack of a better
term, my radio conscience. I am no radio expertwhat I know about radio, I learned because of my
interest in 802.11. Malik knew radio technology before 802.11, and I have been privileged to share in
his insight. Gerry Creager of Texas A&M offered insight into the FCC rules and regulations for
unlicensed devices, which was valuable because wireless LANs have been upending the rules in
recent years. When Glenn Fleishman agreed to write the foreword, I had no idea that he would offer
so much help in placing 802.11 within its larger context. Many of the details he suggested were
references to articles that had run in the past years on his own Wi-Fi Networking News site. As a
writer himself, Glenn also pointed out several locations where better examples would make my points
much clearer. Finally, Terry Simons of the Open1X project has worked extensively with 802.11 on
Linux, and with nearly every 802.1X supplicant on the major operating systems. Terry also is one of
the architects of the wireless authentication system at the University of Utah. His expertise can be
felt throughout the early part of the book on security specifications, as well as in the practical matter
of using supplicants and building an authentication system.
I am also indebted to many others who help keep me abreast of current developments in 802.11,
and share their knowledge with me. Since 2002, I have been privileged to participate in the Interop
Labs initiatives related to wireless security and 802.1X. The real world is far too messy for the
classroom. Every year, I learn more about the state of the art by volunteering than I ever could by
taking a prepared class. Through the Interop Labs, I met Chris Hessing, the development lead for
xsupplicant. Chris has always generously explained how all the keying bits move around in 802.11,
which is no small feat! Sudheer Matta, a colleague of mine, always has time to explain what is
happening in the standards world, and how the minute details of the MAC work.
The large supporting cast at O'Reilly was tremendously helpful in a wide variety of ways. Ellie
Volckhausen designed a stunning cover that has adorned my cubicle as well as most of the personal
electronics devices I own since 2001, when I began writing the first edition. (It even looks good as
the wallpaper on my mobile telephone!) Jessamyn Read took a huge mass of raw sketches and
converted every last one into something that is worth looking at, and did so on a grueling schedule. I
do not know how many hours Colleen Gorman, the production editor, put into this book to get it
finished, but I hope her family and her cat, Phineas, forgive me. And, as always, I am thankful for
the wisdom of Mike Loukides, the editor. Mike kept this project moving forward in the innumerable
ways I have been accustomed to from our past collaborations, and his background as a ham radio
operator proved especially useful when I started writing about the dark and forbidding world of
antennas and RF transmission. (Among many, many other items, you have him to thank for the
footnote on the gain of the Aricebo radio telescope!)
As with so much in life, the devil of writing is in the details. Getting it right means rewriting, and then
probably rewriting some more. I did not attempt a large writing project until college, when I took
Brad Bateman's U.S. Financial System class. Although I certainly learned about the flow of money
through the economy and the tools that the Federal Reserve uses in formulating policy, what I most
value in retrospect was the highly structured process of writing a lengthy paper throughout the
semester. In addition to simply producing a large document, Dr. Bateman stressed the revision
process, a skill that I had to use repeatedly in the preparation of this book and its second edition. It
would be a mistake, however, for me to simply credit Dr. Bateman as an outstanding writing teacher
or an economist gifted with the ability to explain complex subjects to his students. Dr. Bateman is
not shackled by his narrow academic expertise. During the preparation of the second edition of this
book, I attended a lecture of his about the social history of my alma mater. In a captivating hour, he
traced the history of the institution and its intersection with wider social movements, which explained
its present-day culture in far more depth than I ever appreciated while a student. Not all professors
teach to prepare students for graduate school, and not all professors confine their teaching to the
classroom. I am a far better writer, economist, and citizen for his influence.
When writing a book, it is easy to acknowledge the tangible contributions of others. Behind every
author, though, there is a supportive cast of relatives and friends. As always, my wife Ali continued
to indulge my writing habit with extremely good humor, especially considering the number of
weekends that were sacrificed to this book. Many of my friends informally supported this project with
a great deal of encouragement and support; my thanks must go to (in alphabetical order) Annie,
Aramazd, Brian, Dameon, Kevin, and Nick.
Matthew Gast
San Francisco, California
February 2005
Chapter 1. Introduction to Wireless
Networking
Over the past five years, the world has become increasingly mobile. As a result, traditional ways of
networking the world have proven inadequate to meet the challenges posed by our new collective
lifestyle. If users must be connected to a network by physical cables, their movement is dramatically
reduced. Wireless connectivity, however, poses no such restriction and allows a great deal more free
movement on the part of the network user. As a result, wireless technologies are encroaching on the
traditional realm of "fixed" or "wired" networks. This change is obvious to anybody who drives on a
regular basis. One of the "life and death" challenges to those of us who drive on a regular basis is the
daily gauntlet of erratically driven cars containing mobile phone users in the driver's seat.
Wireless connectivity for voice telephony has created a whole new industry. Adding mobile
connectivity into the mix for telephony has had profound influences on the business of delivering
voice calls because callers could be connected to people, not devices. We are on the cusp of an
equally profound change in computer networking. Wireless telephony has been successful because it
enables people to connect with each other regardless of location. New technologies targeted at
computer networks promise to do the same for Internet connectivity. The most successful wireless
data networking technology this far has been 802.11.
In the first edition of this book, I wrote about 802.11 being the tip of the trend in mobile data
networking. At the time, 802.11 and third-generation mobile technologies were duking it out for
mindshare, but 802.11 has unquestionably been more successful to date.
Why Wireless?
To dive into a specific technology at this point is getting a bit ahead of the story, though. Wireless
networks share several important advantages, no matter how the protocols are designed, or even
what type of data they carry.
The most obvious advantage of wireless networking is mobility. Wireless network users can connect
to existing networks and are then allowed to roam freely. A mobile telephone user can drive miles in
the course of a single conversation because the phone connects the user through cell towers.
Initially, mobile telephony was expensive. Costs restricted its use to highly mobile professionals such
as sales managers and important executive decision makers who might need to be reached at a
moment's notice regardless of their location. Mobile telephony has proven to be a useful service,
however, and now it is relatively common in the United States and extremely common among
Europeans.[*]
[*]
While most of my colleagues, acquaintances, and family in the U.S. have mobile telephones, it is still possible to be a holdout.
In Europe, it seems as if everybody has a mobile phoneone cab driver in Finland I spoke with while writing the first edition of this
book took great pride in the fact that his family of four had six mobile telephones!
Likewise, wireless data networks free software developers from the tethers of an Ethernet cable at a
desk. Developers can work in the library, in a conference room, in the parking lot, or even in the
coffee house across the street. As long as the wireless users remain within the range of the base
station, they can take advantage of the network. Commonly available equipment can easily cover a
corporate campus; with some work, more exotic equipment, and favorable terrain, you can extend
the range of an 802.11 network up to a few miles.
Wireless networks typically have a great deal of flexibility, which can translate into rapid deployment.
Wireless networks use a number of base stations to connect users to an existing network. (In an
802.11 network, the base stations are called access points.) The infrastructure side of a wireless
network, however, is qualitatively the same whether you are connecting one user or a million users.
To offer service in a given area, you need base stations and antennas in place. Once that
infrastructure is built, however, adding a user to a wireless network is mostly a matter of
authorization. With the infrastructure built, it must be configured to recognize and offer services to
the new users, but authorization does not require more infrastructure. Adding a user to a wireless
network is a matter of configuring the infrastructure, but it does not involve running cables, punching
down terminals, and patching in a new jack.[ ]
[
] This simple example ignores the challenges of scale. Naturally, if the new users will overload the existing infrastructure, the
infrastructure itself will need to be beefed up. Infrastructure expansion can be expensive and time-consuming, especially if it
involves legal and regulatory approval. However, my basic point holds: adding a user to a wireless network can often be reduced
to a matter of configuration (moving or changing bits) while adding a user to a fixed network requires making physical connections
(moving atoms), and moving bits is easier than moving atoms.
Flexibility is an important attribute for service providers. One of the markets that many 802.11
equipment vendors have been chasing is the so-called "hot spot" connectivity market. Airports and
train stations are likely to have itinerant business travelers interested in network access during
connection delays. Coffeehouses and other public gathering spots are social venues in which network
access is desirable. Many cafes already offer Internet access; offering Internet access over a wireless
network is a natural extension of the existing Internet connectivity. While it is possible to serve a
fluid group of users with Ethernet jacks, supplying access over a wired network is problematic for
several reasons. Running cables is time-consuming and expensive and may also require construction.
Properly guessing the correct number of cable drops is more an art than a science. With a wireless
network, though, there is no need to suffer through construction or make educated (or wild) guesses
about demand. A simple wired infrastructure connects to the Internet, and then the wireless network
can accommodate as many users as needed. Although wireless LANs have somewhat limited
bandwidth, the limiting factor in networking a small hot spot is likely to be the cost of WAN
bandwidth to the supporting infrastructure.
Flexibility may be particularly important in older buildings because it reduces the need for
construction. Once a building is declared historical, remodeling can be particularly difficult. In
addition to meeting owner requirements, historical preservation agencies must be satisfied that new
construction is not desecrating the past. Wireless networks can be deployed extremely rapidly in such
environments because there is only a small wired network to install.
Flexibility has also led to the development of grassroots community networks. With the rapid price
erosion of 802.11 equipment, bands of volunteers are setting up shared wireless networks open to
visitors. Community networks are also extending the range of Internet access past the limitations for
DSL into communities where high-speed Internet access has been only a dream. Community
networks have been particularly successful in out-of-the way places that are too rugged for
traditional wireline approaches.
Like all networks, wireless networks transmit data over a network medium. The medium is a form of
electromagnetic radiation.[*] To be well-suited for use on mobile networks, the medium must be able
to cover a wide area so clients can move throughout a coverage area. Early wireless networks used
infrared light. However, infrared light has limitations; it is easily blocked by walls, partitions, and
other office construction. Radio waves can penetrate most office obstructions and offer a wider
coverage range. It is no surprise that most, if not all, 802.11 products on the market use the radio
wave physical layer.
[*]
Laser light is also used by some wireless networking applications, but the extreme focus of a laser beam makes it suited only
for applications in which the ends are stationary. "Fixed wireless" applications, in which lasers replace other access technology
such as leased telephone circuits, are a common application.
Radio Spectrum: The Key Resource
Wireless devices are constrained to operate in a certain frequency band. Each band has an associated
bandwidth, which is simply the amount of frequency space in the band. Bandwidth has acquired a
connotation of being a measure of the data capacity of a link. A great deal of mathematics,
information theory, and signal processing can be used to show that higher-bandwidth slices can be
used to transmit more information. As an example, an analog mobile telephony channel requires a
20-kHz bandwidth. TV signals are vastly more complex and have a correspondingly larger bandwidth
of 6 MHz.
Early Adoption of 802.11
802.11's explosive advance has not been even. Some markets have evolved more
quickly than others because the value of wireless networks is more pronounced in some
markets. In general, the higher the value placed on mobility and flexibility, the greater
the interest in wireless LANs.
Logistics organizations responsible for moving goods around (think UPS, FedEx, or
airlines), were perhaps the earliest adopters of 802.11. Well before the advent of
802.11, package tracking was done with proprietary wireless LANs. Standardized
products lowered the price and enabled competition between suppliers of network
equipment, and it was an easy decision to replace proprietary products with standardized
ones.
Health care has been an early adopter of wireless networks because of the great
flexibility that is often required of health care equipment. Patients can be moved
throughout a hospital, and the health care professionals that spend time with patients
are among some of the most mobile workers in the economy. Technologically advanced
health care organizations have adopted wireless LANs to make patient information
available over wireless LANs to improve patient care by making information more
accessible to doctors. Computerized records can be transferred between departments
without the requirement to decipher the legendarily illegible doctor scrawls. In the
cluttered environments of an emergency room, rapid access to imaging data can quite
literally be a lifesaver. Several hospitals have deployed PCs to make radiology images
available over wireless LANs on specially-equipped "crash carts" that offer instant access
to X-rays, allowing doctors to make quick decisions without waiting for film to be
developed.
Many eductional institutions have enthusiastically adopted wireless LANs. 10 years ago,
colleges competed for students based on how "wired" the campus was. More high speed
data ports everywhere was assumed to be better. Nowadays, the leading stories in
education are the colleges using wireless LANs to blanket coverage throughout the
campus. Students are highly mobile network users, and can benefit greatly from network
access between classes or in their "homes away from home" (the library, studio, or
science lab, depending on major).
Radio spectrum allocation is rigorously controlled by regulatory authorities through licensing
processes. Most countries have their own regulatory bodies, though regional regulators do exist. In
the U.S., regulation is done by the Federal Communications Commission (FCC). Many FCC rules are
adopted by other countries throughout the Americas. European allocation is performed by the
European Radiocommunications Office (ERO). Other allocation work is done by the International
Telecommunications Union (ITU). To prevent overlapping uses of the radio waves, frequency is
allocated in bands, which are simply ranges of frequencies available to specified applications. Table
1-1 lists some common frequency bands used in the U.S.[*]
[*]
The full spectrum allocation map is available from the National Telecommunications and Information Administration at
http://www.ntia.doc.gov/osmhome/allochrt.pdf.
Table 1-1. Common U.S. frequency bands
Band
Frequency range
UHF ISM
902-928 MHz
S-Band
2-4 GHz
S-Band ISM
2.4-2.5 GHz
C-Band
4-8 GHz
C-Band satellite downlink
3.7-4.2 GHz
C-Band Radar (weather)
5.25-5.925 GHz
C-Band ISM
5.725-5.875 GHz
C-Band satellite uplink
5.925-6.425 GHz
X-Band
8-12 GHz
X-Band Radar (police/weather)
8.5-10.55 GHz
Ku-Band
12-18 GHz
Ku-Band Radar (police)
13.4-14 GHz 15.7-17.7 GHz
The ISM bands
In Table 1-1, there are three bands labeled ISM, which is an abbreviation for industrial, scientific, and
medical. ISM bands are set aside for equipment that, broadly speaking, is related to industrial or
scientific processes or is used by medical equipment. Perhaps the most familiar ISM-band device is
the microwave oven, which operates in the 2.4-GHz ISM band because electromagnetic radiation at
that frequency is particularly effective for heating water.
I pay special attention to the ISM bands in the table because those bands allow license-free
operation, provided the devices comply with power constraints. 802.11 operates in the ISM bands,
along with many other devices. Common cordless phones operate in the ISM bands as well. 802.11b
and 802.11g devices operate within the 2.4 GHz ISM band, while 802.11a devices operate in the 5
GHz band.
The more common 802.11b/g devices operate in S-band ISM. The ISM bands are generally licensefree, provided that devices are low-power. How much sense does it make to require a license for
microwave ovens, after all? Likewise, you don't need a license to set up and operate a low-power
wireless LAN.
What Makes Wireless Networks Different
Wireless networks are an excellent complement to fixed networks, but they are not a replacment
technology. Just as mobile telephones complement fixed-line telephony, wireless LANs complement
existing fixed networks by providing mobility to users. Servers and other data center equipment must
access data, but the physical location of the server is irrelevant. As long as the servers do not move,
they may as well be connected to wires that do not move. At the other end of the spectrum, wireless
networks must be designed to cover large areas to accommodate fast-moving clients. Typical 802.11
access points do not cover large areas, and would have a hard time coping with users on rapidlymoving vehicles.
Lack of Physical Boundary
Traditional network security places a great deal of emphasis on physical security of the network
components. Data on the network travels over well-defined pathways, usually of copper or fiber, and
the network infrastructure is protected by strong physical access control. Equipment is safely locked
away in wiring closets, and set up so that it cannot be reconfigured by users. Basic security stems
from the (admittedly marginal) security of the physical layer. Although it is possible to tap or redirect
signals, physical access control makes it much harder for an intruder to gain surreptitious access to
the network.
Wireless networks have a much more open network medium. By definition, the network medium in a
wireless network is not a well-defined path consisting of a physical cable, but a radio link with a
particular encoding and modulation. Signals can be sent or received by anybody in possession of the
radio techniques, which are of course well known because they are open standards. Interception of
data is child's play, given that the medium is open to anybody with the right network interface, and
the network interface can be purchased for less than $50 at your local consumer electronics store.
Careful shopping online may get you cards for half of that.
Furthermore, radio waves tend to travel outside their intended location. There is no abrupt physical
boundary of the network medium, and the range at which transmissions can be received can be
extended with high-gain antennas on either side. When building a wireless network, you must
carefully consider how to secure the connection to prevent unauthorized use, traffic injection, and
traffic analysis. With the maturation of wireless protocols, the tools to authenticate wireless users
and properly encrypt traffic are now well within reach.
Dynamic Physical Medium
Once a wired network is put in place, it tends to be boring, which is to say, predictable. Once the
cables have been put in place, they tend to do the same thing day in and day out. Provided the
network has been designed according to the engineering rules laid out in the specification, the
network should function as expected. Capacity can be added to a wired network easily by upgrading
the switches in the wiring closet.
In contrast, the physical medium on wireless LANs is much more dynamic. Radio waves bounce off
objects, penetrate through walls, and can often behave somewhat unpredictably. Radio waves can
suffer from a number of propagation problems that may interrupt the radio link, such as multipath
interference and shadows. Without a reliable network medium, wireless networks must carefully
validate received frames to guard against frame loss. Positive acknowledgment, the tactic used by
802.11, does an excellent job at assuring delivery at some cost to throughput.
Radio links are subject to several additional constraints that fixed networks are not. Because radio
spectrum is a relatively scarce resource, it is carefully regulated. Two ways exist to make radio
networks go faster. Either more spectrum can be allocated, or the encoding on the link can be made
more sensitive so that it packs more data in per unit of time. Additional spectrum allocations are
relatively rare, especially for license-free networks. 802.11 networks have kept the bandwidth of a
station's radio channel to approximately 30 MHz, while developing vastly improved encoding to
improve the speed. Faster coding methods can increase the speed, but do have one potential
drawback. Because the faster coding method depends on the receiver to pick out subtle signal
differences, much greater signal-to-noise ratios are required. Higher data rates therefore require the
station to be located closer to its access point. Table 1-2 shows the standardized physical layers in
802.11 and their respective speeds.
Table 1-2. Comparison of 802.11 physical layers (PHYs)
IEEE
standard
Speed
Frequency
band
802.11
1 Mbps 2 2.4 GHz
Mbps
First PHY standard (1997). Featured both frequency-hopping
and direct-sequence modulation techniques.
802.11a
Up to 54 5 GHz
Mbps
Second PHY standard (1999), but products not released until
late 2000.
802.11b
5.5 Mbps 2.4 GHz
11 Mbps
Third PHY standard, but second wave of products. The most
common 802.11 equipment as the first edition of this book was
written, and the majority of the legacy installed base at the
time the second edition was written.
802.11g
Up to 54 2.4 GHz
Mbps
Fourth PHY standard (2003). Applies the coding techniques of
802.11a for higher speed in the 2.4 GHz band, while retaining
backwards compatibility with existing 802.11b networks. The
most common technology included with laptops in 2005.
Notes
Radio is inherently a broadcast medium. When one station transmits, all other stations must listen.
Access points act much like old shared Ethernet hubs in that there is a fixed amount of transmission
capacity per access point, and it must be shared by all the attached users. Adding capacity requires
that the network administrator add access points while simultaneously reducing the coverage area of
existing access points.
Security
Many wireless networks are based on radio waves, which makes the network medium inherently
open to interception. Properly protecting radio transmissions on any network is always a concern for
protocol designers. 802.11 did not build in much in the way of security protocols. Coping with the
inherent unreliability of the wireless medium and mobility required several protocol features to
confirm frame delivery, save power, and offer mobility. Security was quite far down the list, and
proved inadequate in the early specifications.
Wireless networks must be strongly authenticated to prevent use by unauthorized users, and
authenticated connections must be strongly encrypted to prevent traffic interception and injection by
unauthorized parties. Technologies that offer strong encryption and authentication have emerged
since the first edition of this book, and are a major component of the revisions for the second edition.
A Network by Any Other Name...
Wireless networking is a hot industry segment. Several wireless technologies have been targeted
primarily for data transmission. Bluetooth is a standard used to build small networks between
peripherals: a form of "wireless wires," if you will. Most people in the industry are familiar with the
hype surrounding Bluetooth, though it seems to have died down as real devices have been brought
to market. In the first edition, I wrote that I have not met many people who have used Bluetooth
devices, but it is much more common these days. (I use a Bluetooth headset on a regular basis.)
Post-second-generation (2.5G) and third-generation (3G) mobile telephony networks are also a
familiar wireless technology. They promise data rates of megabits per cell, as well as the "always on"
connections that have proven to be quite valuable to DSL and cable modem customers. After many
years of hype and press from 3G equipment vendors, the rollout of commercial 3G services is finally
underway. 2.5G services like GPRS, EDGE, and 1xRTT are now widely available, and third-generation
networks based on UMTS or EV-DO are quickly being built. (I recently subscribed to an unlimited
GPRS service to get connected during my train trips between my office and my home.) Many articles
quote peak speeds for these technologies in the hundreds of kilobits per second or even megabits,
but this capacity must be shared between all users in a cell. Real-world downstream speeds are
roughly comparable to dial-up modem connections and cannot touch an 802.11 hot spot.
This is a book about 802.11 networks. 802.11 goes by a variety of names, depending on who is
talking about it. Some people call 802.11 wireless Ethernet, to emphasize its shared lineage with the
traditional wired Ethernet (802.3). A second name which has grown dramatically in popularity since
the first edition of this book is Wi-Fi, from the interoperability certification program run by the Wi-Fi
Alliance, the major trade assocation of 802.11 equipment vendors. The Wi-Fi Alliance, formerly
known as the Wireless Ethernet Compatibility Alliance (WECA), will test member products for
compatibility with 802.11 standards.[*] Other organizations will perform compatibility testing as well;
the University of New Hampshire's InterOperability Lab (IOL) recently launched a wireless test
consortium.
[*]
More details on the Wi-Fi Alliance and its certification program can be found at http://www.wi-fi.org/.
The Wonderful Thing About Standards...
Several standards groups are involved in 802.11-related standardization efforts because 802.11 cuts
across many formerly distinct boundaries in networking. Most of the effort remains concentrated in
the IEEE, but important contributions to wireless LAN standards have come from several major
locations.
The first is the Institute of Electronics and Electrical Engineers (IEEE). In addition to its activities as a
professional society, the IEEE works on standardizing electrical equipment, including several types of
communication technology. IEEE standardization efforts are organized by projects, each of which is
assigned a number. By far the most famous IEEE project is the IEEE 802 project to develop LAN
standards. Within a project, individual working groups develop standards to address a particular facet
of the problem. Working groups are also given a number, which is written after the decimal point for
the corresponding projects. Ethernet, the most widely used IEEE LAN technology, was standardized
by the third working group, 802.3. Wireless LANs were the eleventh working group formed, hence
the name 802.11.
Within a working group, task groups form to revise particular aspects of the standard or add on to
the general area of functionality. Task groups are assigned a letter beneath the working group, and
the document produced by a task group combines the project and working group number, followed
by the letter from the task group. (Some letters that are subject to easy confusion with letters, such
as the lowercase "l," are not used.) In wireless networking, the first task group to gain wide
recognition was Task Group B (TGb), which produced the 802.11b specification. Table 1-3 is a basic
listing of the different 802.11 standards.
Interestingly enough, the case of the letter in a standards revision encodes
information. Lowercase letters indicate dependent standards that cannot stand
alone from their parent, while uppercase letters indicate full-fledged standalone
specifications.
802.11b adds a new clause to 802.11, but cannot stand alone, so the "b" is
written in lowercase. In constrast, standards like 802.1Q and 802.1X are
standalone specifications that are completely self-contained in one document,
and therefore use uppercase letters.
At periodic intervals, the additions from dependent task groups will be "rolled up" into the main
parent specification. The initial revision of 802.11 came out in 1997. Minor changes to the text were
released as 802.11-1999, which was the baseline standard for quite some time. The most recent
rollup is 802.11-2003.
Table 1-3. standards
IEEE
standard
Notes
802.11
First standard (1997). Specified the MAC and the original slower frequency-hopping
and direct-sequence modulation techniques.
802.11a
Second physical layer standard (1999), but products not released until late 2000.
802.11b
Third physical layer standard (1999), but second wave of products. The most
common 802.11 equipment as the first book was written.
TGc
Task group that produced a correction to the example encoding in 802.11a. Since
the only product was a correction, there is no 802.11c.
802.11d
Extends frequency-hopping PHY for use across multiple regulatory domains
TGe (future
802.11e)
Task group producing quality-of-service (QoS) extensions for the MAC. An interim
snapshot called Wi-Fi Multi-Media (WMM) is likely to be implemented before the
standard is complete.
802.11F
Inter-access point protocol to improve roaming between directly attached access
points
802.11g
Most recently standardized (2003) PHY for networks in the ISM band.
802.11h
Standard to make 802.11a compatible with European radio emissions regulations.
Other regulators have adopted its mechanisms for different purposes.
802.11i
Improvements to security at the link layer.
802.11j
Enhancements to 802.11a to conform to Japanese radio emission regulations.
IEEE
standard
Notes
TGk (future
802.11k)
Task group to enhance communication between clients and network to better
manage scarce radio use.
TGm
Task group to incorporate changes made by 802.11a, 802.11b, and 802.11d, as
well as changes made by TGc into the main 802.11 specification. (Think "m" for
maintenance.)
TGn (future
802.11n)
Task group founded to create a high-throughput standard. The design goal is
throughput in excess of 100 Mbps, and the resulting standard will be called
802.11n.
TGp (future
802.11p)
Task group adopting 802.11 for use in automobiles. The initial use is likely to be a
standard protocol used to collect tolls.
TGr (future
802.11r)
Enhancements to roaming performance.
TGs (future
802.11s)
Task group enhancing 802.11 for use as mesh networking technology.
TGT (future
802.11T)
Task group designing test and measurement specification for 802.11. Its result will
be standalone, hence the uppercase letter.
TGu (future
802.11u)
Task group modifying 802.11 to assist in interworking with other network
technologies.
When it became clear that authentication on wireless networks was fundamentally broken, the IEEE
adopted several authentication standards originally developed by the Internet Engineering Task Force
(IETF). Wireless LAN authentication depends heavily on protocols defined by the IETF.
The Wi-Fi Alliance is a combination of a trade association, testing organization, and standardization
organization. Most of the Wi-Fi Alliance's emphasis is on acting as a trade association for its
members, though it also well-known for the Wi-Fi certification program. Products are tested for
interoperability with a testbed consisting of products from major vendors, and products that pass the
test suite are awarded the right to use the Wi-Fi mark.
The Wi-Fi Alliance's standardization efforts are done in support of the IEEE. When the security of
wireless networks was called into question, the Wi-Fi Alliance produced an interim security
specification called Wi-Fi Protected Access (WPA). WPA was essentially a snapshot of the work done
by the IEEE security task group. It is more of a marketing standard than a technical standard, since
the technology was developed by the IEEE. However, it serves a role in accelerating the development
of secure wireless LAN solutions.
Chapter 2. Overview of 802.11 Networks
Before studying the details of anything, it often helps to get a general "lay of the land." A basic
introduction is often necessary when studying networking topics because the number of acronyms
can be overwhelming. Unfortunately, 802.11 takes acronyms to new heights, which makes the
introduction that much more important. To understand 802.11 on anything more than a superficial
basis, you must get comfortable with some esoteric terminology and a herd of three-letter acronyms.
This chapter is the glue that binds the entire book together. Read it for a basic understanding of
802.11, the concepts that will likely be important to users, and how the protocol is designed to
provide an experience as much like Ethernet as possible. After that, move on to the low-level
protocol details or deployment, depending on your interests and needs.
Part of the reason this introduction is important is because it introduces the acronyms used
throughout the book. With 802.11, the introduction serves another important purpose. 802.11 is
superficially similar to Ethernet. Understanding the background of Ethernet helps slightly with
802.11, but there is a host of additional background needed to appreciate how 802.11 adapts
traditional Ethernet technology to a wireless world. To account for the differences between wired
networks and the wireless media used by 802.11, a number of additional management features were
added. At the heart of 802.11 is a white lie about the meaning of media access control (MAC).
Wireless network interface cards are assigned 48-bit MAC addresses, and, for all practical purposes,
they look like Ethernet network interface cards. In fact, the MAC address assignment is done from
the same address pool so that 802.11 cards have unique addresses even when deployed into a
network with wired Ethernet stations.
To outside network devices, these MAC addresses appear to be fixed, just as in other IEEE 802
networks; 802.11 MAC addresses go into ARP tables alongside Ethernet addresses, use the same set
of vendor prefixes, and are otherwise indistinguishable from Ethernet addresses. The devices that
comprise an 802.11 network (access points and other 802.11 devices) know better. There are many
differences between an 802.11 device and an Ethernet device, but the most obvious is that 802.11
devices are mobile; they can easily move from one part of the network to another. The 802.11
devices on your network understand this and deliver frames to the current location of the mobile
station.
IEEE 802 Network Technology Family Tree
802.11 is a member of the IEEE 802 family, which is a series of specifications for local area network
(LAN) technologies. Figure 2-1 shows the relationship between the various components of the 802
family and their place in the OSI model.
Figure 2-1. The IEEE 802 family and its relation to the OSI model
IEEE 802 specifications are focused on the two lowest layers of the OSI model because they
incorporate both physical and data link components. All 802 networks have both a MAC and a
Physical (PHY) component. The MAC is a set of rules to determine how to access the medium and
send data, but the details of transmission and reception are left to the PHY.
Individual specifications in the 802 series are identified by a second number. For example, 802.3 is
the specification for a Carrier Sense Multiple Access network with Collision Detection (CSMA/CD),
which is related to (and often mistakenly called) Ethernet, and 802.5 is the Token Ring specification.
Other specifications describe other parts of the 802 protocol stack. 802.2 specifies a common link
layer, the Logical Link Control (LLC), which can be used by any lower-layer LAN technology.
Management features for 802 networks are specified in 802.1. Among 802.1's many provisions are
bridging (802.1D) and virtual LANs, or VLANs (802.1Q).
802.11 is just another link layer that can use the 802.2/LLC encapsulation. The base 802.11
specification includes the 802.11 MAC and two physical layers: a frequency-hopping spread-spectrum
(FHSS) physical layer and a direct-sequence spread-spectrum (DSSS) link layer. Later revisions to
802.11 added additional physical layers. 802.11b specifies a high-rate direct-sequence layer
(HR/DSSS); products based on 802.11b hit the marketplace in 1999 and was the first mass-market
PHY. 802.11a describes a physical layer based on orthogonal frequency division multiplexing
(OFDM); products based on 802.11a were released as the first edition of this book was completed.
802.11g is the newest physical layer on the block. It offers higher speed through the use of OFDM,
but with backwards compatibility with 802.11b. Backwards compatibility is not without a price,
though. When 802.11b and 802.11g users coexist on the same access point, additional protocol
overhead is required, reducing the maximum speed for 802.11g users.
To say that 802.11 is "just another link layer for 802.2" is to omit the details in the rest of this book,
but 802.11 is exciting precisely because of these details. 802.11 allows for mobile network access; in
accomplishing this goal, a number of additional features were incorporated into the MAC. As a result,
the 802.11 MAC may seem baroquely complex compared to other IEEE 802 MAC specifications.
The use of radio waves as a physical layer requires a relatively complex PHY, as well. 802.11 splits
the PHY into two generic PMcomponents: the Physical Layer Convergence Procedure (PLCP), to map
the MAC frames onto the medium, and a Physical Medium Dependent (PMD) system to transmit
those frames. The PLCP straddles the boundary of the MAC and physical layers, as shown in Figure 22. In 802.11, the PLCP adds a number of fields to the frame as it is transmitted "in the air."
Figure 2-2. PHY components
All this complexity begs the question of how much you actually need to know. As with any
technology, the more you know, the better off you will be. The 802.11 protocols have many knobs
and dials that you can tweak, but most 802.11 implementations hide this complexity. Many of the
features of the standard come into their own only when the network is congested, either with a lot of
traffic or with a large number of wireless stations. Networks are increasingly pushing the limits in
both respects. At any rate, I can't blame you for wanting to skip the chapters about the protocols and
jump ahead to the chapters about planning and installing an 802.11 network. After you've read this
chapter, you can skip ahead to Chapters 17-23 and return to the chapters on the protocol's inner
workings when you need (or want) to know more.
802.11 Nomenclature and Design
802.11 networks consist of four major physical components, which are summarized in Figure 2-3.
Figure 2-3. Components of 802.11 LANs
The components are:
Stations
Networks are built to transfer data between stations. Stations are computing devices with
wireless network interfaces. Typically, stations are battery-operated laptop or handheld
computers. There is no reason why stations must be portable computing devices, though. In
some environments, wireless networking is used to avoid pulling new cable, and desktops are
connected by wireless LANs. Large open areas may also benefit from wireless networking, such
as a manufacturing floor using a wireless LAN to connect components. 802.11 is fast becoming
a de facto standard for linking together consumer electronics. Apple's AirPort Express connects
computers to stereos via 802.11. TiVos can connect to wireless networks. Several consumer
electronics companies have joined the 802.11 working group, apparently with the intent of
enabling high-speed media transfers over 802.11.
Access points
Frames on an 802.11 network must be converted to another type of frame for delivery to the
rest of the world. Devices called access points perform the wireless-to-wired bridging function.
(Access points perform a number of other functions, but bridging is by far the most important.)
Initially, access point functions were put into standalone devices, though several newer
products are dividing the 802.11 protocol between "thin" access points and AP controllers.
Wireless medium
To move frames from station to station, the standard uses a wireless medium. Several different
physical layers are defined; the architecture allows multiple physical layers to be developed to
support the 802.11 MAC. Initially, two radio frequency (RF) physical layers and one infrared
physical layer were standardized, though the RF layers have proven far more popular. Several
additional RF layers have been standardized as well.
Distribution system
When several access points are connected to form a large coverage area, they must
communicate with each other to track the movements of mobile stations. The distribution
system is the logical component of 802.11 used to forward frames to their destination. 802.11
does not specify any particular technology for the distribution system. In most commercial
products, the distribution system is implemented as a combination of a bridging engine and a
distribution system medium, which is the backbone network used to relay frames between
access points; it is often called simply the backbone network. In nearly all commercially
successful products, Ethernet is used as the backbone network technology.
Types of Networks
The basic building block of an 802.11 network is the basic service set (BSS), which is simply a group
of stations that communicate with each other. Communications take place within a somewhat fuzzy
area, called the basic service area, defined by the propagation characteristics of the wireless
medium.[*] When a station is in the basic service area, it can communicate with the other members
of the BSS. BSSs come in two flavors, both of which are illustrated in Figure 2-4.
[*]
All of the wireless media used will propagate in three dimensions. From that perspective, the service area should perhaps be
called the service volume. However, the term area is widely used and accepted.
Figure 2-4. Independent and infrastructure BSSs
Independent networks
On the left is an independent BSS (IBSS). Stations in an IBSS communicate directly with each other
and thus must be within direct communication range. The smallest possible 802.11 network is an
IBSS with two stations. Typically, IBSSs are composed of a small number of stations set up for a
specific purpose and for a short period of time. One common use is to create a short-lived network to
support a single meeting in a conference room. As the meeting begins, the participants create an
IBSS to share data. When the meeting ends, the IBSS is dissolved.[ ] Due to their short duration,
small size, and focused purpose, IBSSs are sometimes referred to as ad hoc BSSs or ad hoc
networks.
[
]
IBSSs have found a similar use at LAN parties throughout the world.
Infrastructure networks
On the right side of Figure 2-4 is an infrastructure BSS. (To avoid overloading the acronym, an
infrastructure BSS is never called an IBSS). Infrastructure networks are distinguished by the use of
an access point. Access points are used for all communications in infrastructure networks, including
communication between mobile nodes in the same service area. If one mobile station in an
infrastructure BSS needs to communicate with a second mobile station, the communication must
take two hops. First, the originating mobile station transfers the frame to the access point. Second,
the access point transfers the frame to the destination station. With all communications relayed
through an access point, the basic service area corresponding to an infrastructure BSS is defined by
the points in which transmissions from the access point can be received. Although the multihop
transmission takes more transmission capacity than a directed frame from the sender to the receiver,
it has two major advantages:
An infrastructure BSS is defined by the distance from the access point. All mobile stations are
required to be within reach of the access point, but no restriction is placed on the distance
between mobile stations themselves. Allowing direct communication between mobile stations
would save transmission capacity but at the cost of increased physical layer complexity because
mobile stations would need to maintain neighbor relationships with all other mobile stations
within the service area.
Access points in infrastructure networks are in a position to assist with stations attempting to
save power. Access points can note when a station enters a power-saving mode and buffer
frames for it. Battery-operated stations can turn the wireless transceiver off and power it up
only to transmit and retrieve buffered frames from the access point.
In an infrastructure network, stations must associate with an access point to obtain network services.
Association is the process by which mobile station joins an 802.11 network; it is logically equivalent
to plugging in the network cable on an Ethernet. It is not a symmetric process. Mobile stations
always initiate the association process, and access points may choose to grant or deny access based
on the contents of an association request. Associations are also exclusive on the part of the mobile
station: a mobile station can be associated with only one access point.[*] The 802.11 standard places
no limit on the number of mobile stations that an access point may serve. Implementation
considerations may, of course, limit the number of mobile stations an access point may serve. In
practice, however, the relatively low throughput of wireless networks is far more likely to limit the
number of stations placed on a wireless network.
[*]
One reviewer noted that a similar restriction was present in traditional Ethernet networks until the development of VLANs and
specifically asked how long this restriction was likely to last. I am not intimately involved with the standardization work, so I cannot
speak to the issue directly. I do, however, agree that it is an interesting question.
Extended service areas
BSSs can create coverage in small offices and homes, but they cannot provide network coverage to
larger areas. 802.11 allows wireless networks of arbitrarily large size to be created by linking BSSs
into an extended service set (ESS). An ESS is created by chaining BSSs together with a backbone
network. All the access points in an ESS are given the same service set identifier (SSID), which
serves as a network "name" for the users.
802.11 does not specify a particular backbone technology; it requires only that the backbone provide
a specified set of services. In Figure 2-5, the ESS is the union of the four BSSs (provided that all the
access points are configured to be part of the same ESS). In real-world deployments, the degree of
overlap between the BSSs would probably be much greater than the overlap in Figure 2-5. In real
life, you would want to offer continuous coverage within the extended service area; you wouldn't
want to require that users walk through the area covered by BSS3 when en route from BSS1 to
BSS2.
Figure 2-5. Extended service set
Stations within the same ESS may communicate with each other, even though these stations may be
in different basic service areas and may even be moving between basic service areas. For stations in
an ESS to communicate with each other, the wireless medium must act like a single layer 2
connection. Access points act as bridges, so direct communication between stations in an ESS
requires that the backbone network also look like a layer 2 connection. First-generation access points
required direct layer 2 connections through hubs or virtual LANs; newer products implement a variety
of tunneling technologies to emulate the layer 2 environment.
802.11 supplies link-layer mobility within an ESS, but only if the backbone
network appears to be a single link-layer domain. This important constraint on
mobility is often a major factor in the way that wireless LANs are deployed, and
one of the major ways that vendors differentiate their products.
Early access points required that the backbone network be a single hub or
VLAN, but newer products can interface directly with the backbone. Many can
support multiple VLANs simultaneously with 802.1Q tags, and some can even
dynamically instantiate VLANs based on authentication information.
Extended service areas are the highest-level abstraction supported by 802.11 networks. Access
points in an ESS operate in concert to allow the outside world to use the station's MAC address to
talk to a station no matter what its location is within the ESS. In Figure 2-5, the router uses the
station's MAC address as the destination to deliver frames to a mobile station; only the access point
with which that mobile station is associated delivers the frame. The router remains ignorant of the
location of the mobile station and relies on the access points to deliver the frame.
Multi-BSS environments: "virtual APs"
Early 802.11 radio chips had the ability to create a single basic service set. An AP could have connect
users to only one "wireless network," and all users on that network had similar, if not identical,
privileges. In early deployments with limited user counts, a single logical network was sufficient. As
wireless networking grew in popularity, one network no longer sufficed.
As an example, most organizations get regular visitors, many of whom have 802.11 equipment and
need (or strongly desire) Internet access. Guests are not trusted users. One common way of coping
with guest access is to create two extended service sets on the same physical infrastructure. Current
802.11 chipsets can create multiple networks with the same radio. Using modern chipsets, each
access point hardware device can create two BSSs, one for the network named guest, and one for
the network named internal. Within the AP, each SSIDs is associated with a VLAN. The guest network
is connected to a VLAN prepared for public access by unknown and untrusted users, and is almost
certainly attached outside the firewall.
Wireless devices see two separate networks in the radio domain, and can connect to whatever one
suits their needs. (Naturally, the internal network is probably protected by authentication prevent
unauthorized use.) Users who connect to the wireless network named guest will be placed on the
guest VLAN, while users who connect to the wireless network named internal will be authenticated
and placed on the internal network.
This somewhat contrived example illustrates the development of what many call virtual access
points. Each BSS acts like its own self-contained AP, with its own ESSID, MAC address,
authentication configuration, and encryption settings. Virtual APs are also used to create parallel
networks with different security levels, a configuration that will be discussed in Chapter 22. Current
802.11 radio chipsets have the ability to create 32 or even 64 BSSes, which is adequate for nearly
every configuration.
Robust security networks (RSNs)
Early wireless LANs proved to have feeble built-in security. 802.11i, which was ratified in June 2004,
specifies a set of improved security mechanisms that provide robust security network associations
(RSNAs). Robust security network associations are formed when improved the authentication and
confidentiality protocols defined in 802.11i are in use. Support for 802.11i may be composed of
hardware, software, or both, depending on the exact architecture of a particular device. Hardware
which does not support the improved protocols is referred to as pre-RSN capable. Many recent preRSN capable devices may be upgradeable to support 802.11i, but most older devices will not be
upgradeable.
The Distribution System, Revisited
With an understanding of how an extended service set is built, I'd like to return to the concept of the
distribution system. 802.11 describes the distribution system in terms of the services it provides to
wireless stations. While these services will be described in more detail later in this chapter, it is worth
describing their operation at a high level. The distribution system provides mobility by connecting
access points. When a frame is given to the distribution system, it is delivered to the right access
point and relayed by that access point to the intended destination.
The distribution system is responsible for tracking where a station is physically located and delivering
frames appropriately. When a frame is sent to a mobile station, the distribution system is charged
with the task of delivering it to the access point serving the mobile station. As an example, consider
the router in Figure 2-5. The router simply uses the MAC address of a mobile station as its
destination. The distribution system of the ESS pictured in Figure 2-5 must deliver the frame to the
right access point. Obviously, part of the delivery mechanism is the backbone Ethernet, but the
backbone network cannot be the entire distribution system because it has no way of choosing
between access points. In the language of 802.11, the backbone Ethernet is the distribution system
medium, but it is not the entire distribution system.
To find the rest of the distribution system, we need to look to the access points themselves. Most
access points currently on the market operate as bridges. They have at least one wireless network
interface and at least one Ethernet network interface. The Ethernet side can be connected to an
existing network, and the wireless side becomes an extension of that network. Relaying frames
between the two network media is controlled by a bridging engine.
Figure 2-6 illustrates the relationship between the access point, the backbone network, and the
distribution system. The access point has two interfaces connected by a bridging engine. Arrows
indicate the potential paths to and from the bridging engine. Frames may be sent by the bridge to
the wireless network; any frames sent by the bridge's wireless port are transmitted to all associated
stations. Each associated station can transmit frames to the access point. Finally, the backbone port
on the bridge can interact directly with the backbone network. The distribution system in Figure 2-6
is composed of the bridging engine plus the wired backbone network.
Figure 2-6. Distribution system in common 802.11 access point
implementations
Every frame sent by a mobile station in an infrastructure network must use the distribution system.
It is easy to understand why interaction with hosts on the backbone network must use the
distribution system. After all, they are connected to the distribution system medium. Wireless
stations in an infrastructure network depend on the distribution system to communicate with each
other because they are not directly connected to each other. The only way for station A to send a
frame to station B is by relaying the frame through the bridging engine in the access point. However,
the bridge is a component of the distribution system. While what exactly makes up the distribution
system may seem like a narrow technical concern, there are some features of the 802.11 MAC that
are closely tied to its interaction with the distribution system.
Interaccess point communication as part of the distribution system
Included with this distribution system is a method to manage associations. A wireless station is
associated with only one access point at a time. If a station is associated with one access point, all
the other access points in the ESS need to learn about that station. In Figure 2-5, AP4 must know
about all the stations associated with AP1. If a wireless station associated with AP4 sends a frame to
a station associated with AP1, the bridging engine inside AP4 must send the frame over the backbone
Ethernet to AP1 so it can be delivered to its ultimate destination. To fully implement the distribution
system, access points must inform other access points of associated stations. Naturally, many access
points on the market use an interaccess point protocol (IAPP) over the backbone medium. Many
vendors developed proprietary protocols between access points to carry association data. A standard
IAPP was produced as 802.11F, but I am not aware of its use in any products.
Wireless bridges and the distribution system
Up to this point, I have tacitly assumed that the distribution system medium was an existing fixed
network. While this will often be the case, the 802.11 specification explicitly supports using the
wireless medium itself as the distribution system. The wireless distribution system (WDS)
configuration is often called a "wireless bridge" configuration because it allows network engineers to
connect two LANs at the link layer. Wireless bridges can be used to quickly connect distinct physical
locations and are well-suited for use by access providers. Most 802.11 access points on the market
now support the wireless bridge configuration, though it may be necessary to upgrade the firmware
on older units.
Network Boundaries
Because of the nature of the wireless medium, 802.11 networks have fuzzy boundaries. In fact, some
degree of fuzziness is desirable. As with mobile telephone networks, allowing basic service areas to
overlap increases the probability of successful transitions between basic service areas and offers the
highest level of network coverage. The basic service areas on the right of Figure 2-7 overlap
significantly. This means that a station moving from BSS2 to BSS4 is not likely to lose coverage; it
also means that AP3 (or, for that matter, AP4) can fail without compromising the network too badly.
On the other hand, if AP2 fails, the network is cut into two disjoint parts, and stations in BSS1 lose
connectivity when moving out of BSS1 and into BSS3 or BSS4. Coping with "coverage holes" from
access point failures is a task that requires attention during the network design phase; many newer
products offer dynamic radio tuning capabilities to automatically fill in holes that develop during
network operation.
Figure 2-7. Overlapping BSSs in an ESS
Different types of 802.11 networks may also overlap. Independent BSSs may be created within the
basic service area of an access point. Figure 2-8 illustrates spatial overlap. An access point appears
at the top of the figure; its basic service area is shaded. Two stations are operating in infrastructure
mode and communicate only with the access point. Three stations have been set up as an
independent BSS and communicate with each other. Although the five stations are assigned to two
different BSSs, they may share the same wireless medium. Stations may obtain access to the
medium only by using the rules specified in the 802.11 MAC; these rules were carefully designed to
enable multiple 802.11 networks to coexist in the same spatial area. Both BSSs must share the
capacity of a single radio channel, so there may be adverse performance implications from co-located
BSSs.
Figure 2-8. Overlapping network types
802.11 Network Operations
From the outset, 802.11 was designed to be just another link layer to higher-layer protocols.
Network administrators familiar with Ethernet will be immediately comfortable with 802.11. The
shared heritage is deep enough that 802.11 is sometimes referred to as "wireless Ethernet."
The core elements present in Ethernet are present in 802.11. Stations are identified by 48-bit IEEE
802 MAC addresses. Conceptually, frames are delivered based on the MAC address. Frame delivery is
unreliable, though 802.11 incorporates some basic reliability mechanisms to overcome the inherently
poor qualities of the radio channels it uses.[*]
[*]
I don't mean "poor" in an absolute sense. But the reliability of wireless transmission is really not comparable to the reliability of a
wired network.
From a user's perspective, 802.11 might just as well be Ethernet. Network administrators, however,
need to be conversant with 802.11 at a much deeper level. Providing MAC-layer mobility while
following the path blazed by previous 802 standards requires a number of additional services and
more complex framing.
Network Services
One way to define a network technology is to define the services it offers and allow equipment
vendors to implement those services in whatever way they see fit. 802.11 provides nine services.
Only three of the services are used for moving data; the remaining six are management operations
that allow the network to keep track of the mobile nodes and deliver frames accordingly.
The services are described in the following list and summarized in Table 2-1:
Distribution
This service is used by mobile stations in an infrastructure network every time they send data.
Once a frame has been accepted by an access point, it uses the distribution service to deliver
the frame to its destination. Any communication that uses an access point travels through the
distribution service, including communications between two mobile stations associated with the
same access point.
Integration
Integration is a service provided by the distribution system; it allows the connection of the
distribution system to a non-IEEE 802.11 network. The integration function is specific to the
distribution system used and therefore is not specified by 802.11, except in terms of the
services it must offer.
Association
Delivery of frames to mobile stations is made possible because mobile stations register, or
associate, with access points. The distribution system can then use the registration information
to determine which access point to use for any mobile station. Unassociated stations are not
"on the network," much like workstations with unplugged Ethernet cables. 802.11 specifies the
function that must be provided by the distribution system using the association data, but it
does not mandate any particular implementation. When robust security network protocols are
in use, association is a precursor to authentication. Prior to the completion of authentication,
an access point will drop all network protocol traffic from a station.
Reassociation
When a mobile station moves between basic service areas within a single extended service
area, it must evaluate signal strength and perhaps switch the access point with which it is
associated. Reassociations are initiated by mobile stations when signal conditions indicate that
a different association would be beneficial; they are never initiated directly by the access point.
(Some APs will kick stations off in order to force a client into being the reassociation process;
in the future, reassociation may be more dependent on the infrastructure with the development
of better network management standards.)
After the reassociation is complete, the distribution system updates its location records to
reflect the reachability of the mobile station through a different access point. As with the
association service, a robust security network will drop network protocol traffic before the
successful completion of authentication.
Disassociation
To terminate an existing association, stations may use the disassociation service. When
stations invoke the disassociation service, any mobility data stored in the distribution system is
removed. Once disassociation is complete, it is as if the station is no longer attached to the
network. Disassociation is a polite task to do during the station shutdown process. The MAC is,
however, designed to accommodate stations that leave the network without formally
disassociating.
Authentication
Physical security is a major component of a wired LAN security solution. Network attachment
points are limited, often to areas in offices behind perimeter access control devices. Network
equipment can be secured in locked wiring closets, and data jacks in offices and cubicles can
be connected to the network only when needed. Wireless networks cannot offer the same level
of physical security, however, and therefore must depend on additional authentication routines
to ensure that users accessing the network are authorized to do so. Authentication is a
necessary prerequisite to association because only authenticated users are authorized to use
the network.
Authentication may happen multiple times during the connection of a client to a wireless
network. Prior to association, a station will perform a basic identity exchange with an access
point consisting of its MAC address. This exchange is often referred to as "802.11"
authentication, which is distinct from the robust cryptographic user authentication that often
follows.
Deauthentication
Deauthentication terminates an authenticated relationship. Because authentication is needed
before network use is authorized, a side effect of deauthentication is termination of any current
association. In a robust security network, deauthentication also clears keying information.
Confidentiality
Strong physical controls can prevent a great number of attacks on the privacy of data in a
wired LAN. Attackers must obtain physical access to the network medium before attempting to
eavesdrop on traffic. On a wired network, physical access to the network cabling is a subset of
physical access to other computing resources. By design, physical access to wireless networks
is a comparatively simpler matter of using the correct antenna and modulation methods.
In the initial revision of 802.11, the confidentiality service was called privacy, and provided by
the now-discredited Wired Equivalent Privacy (WEP) protocol. In addition to new encryption
schemes, 802.11i augments the confidentiality service by providing user-based authentication
and key management services, two critical issues that WEP failed to address.
MSDU delivery
Networks are not much use without the ability to get the data to the recipient. Stations provide
the MAC Service Data Unit (MSDU) delivery service, which is responsible for getting the data to
the actual endpoint.
Transmit Power Control (TPC)
TPC is a new service that was defined by 802.11h. European standards for the 5 GHz band
require that stations control the power of radio transmissions to avoid interfering with other
users of the 5 GHz band. Transmit power control also helps avoid interference with other
wireless LANs. Range is a function of power; high transmit power settings make it more likely
that a client's greater range will interfere with a neighboring network. By controlling power to a
level that is "just right," it is less likely that a station will interfere with neighboring stations.
Dynamic Frequency Selection (DFS)
Some radar systems operate in the 5 GHz range. As a result, some regulatory authorities have
mandated that wireless LANs must detect radar systems and move to frequencies that are not
in use by radar. Some regulatory authorities also require uniform use of the 5 GHz band for
wireless LANs, so networks must have the ability to re-map channels so that usage is
equalized.
Table 2-1. Network services
Service
Station or distribution
Description
service?
Distribution
Distribution
Service used in frame delivery to determine
destination address in infrastructure networks
Integration
Distribution
Frame delivery to an IEEE 802 LAN outside the
wireless network
Association
Distribution
Used to establish the AP which serves as the
gateway to a particular mobile station
Reassociation
Distribution
Used to change the AP which serves as the
gateway to a particular mobile station
Disassociation
Distribution
Removes the wireless station from the network
Authentication
Station
Establishes station identity (MAC address) prior
to establishing association
Deauthentication
Station
Used to terminate authentication, and by
extension, association
Confidentiality
Station
Provides protection against eavesdropping
MSDU delivery
Station
Delivers data to the recipient
Transmit Power Control
(TPC)
Station/spectrum
management
Reduces interference by minimizing station
transmit power
Dynamic Frequency
Selection (DFS)
Station/spectrum
management
Avoids interfering with radar operation in the 5
GHz band
Station services
Station services are part of every 802.11-compliant station and must be incorporated by any product
claiming 802.11 compliance. Station services are provided by both mobile stations and the wireless
interface on access points. Stations provide frame delivery services to allow message delivery, and,
in support of this task, they may need to use the authentication services to establish associations.
Stations may also wish to take advantage of confidentiality functions to protect messages as they
traverse the vulnerable wireless link.
Distribution system services
Distribution system services connect access points to the distribution system. The major role of
access points is to extend the services on the wired network to the wireless network; this is done by
providing the distribution and integration services to the wireless side. Managing mobile station
associations is the other major role of the distribution system. To maintain association data and
station location information, the distribution system provides the association, reassociation, and
disassociation services.
Confidentiality and access control
Confidentiality and access control services are intertwined. In addition to secrecy of the data in
transit, the confidentiality service also proves the integrity of frame contents. Both secrecy and
integrity depend on shared cryptographic keying, so the confidentiality service necessarily depends
on other services to provide authentication and key management.
Authentication and key management (AKM)
Cryptographic integrity is worthless if it does not prevent unauthorized users from attaching to
the network. The confidentiality service depends on the authentication and key management
suite to establish user identity and encryption keys. Authentication may be accomplished
through an external protocol, such as 802.1X, or with pre-shared keys.
Cryptographic algorithms
Frames may be protected by the traditional WEP algorithm, using 40- or 104-bit secret keys,
the Temporal Key Integrity Protocol (TKIP), or the Counter Mode CBC-MAC Protocol (CCMP). All
of these algorithms are discussed in detail in Chapters 5 and 7.
Origin authenticity
TKIP and CCMP allow the receiver to validate the sender's MAC address to prevent spoofing
attacks. Origin authenticity protection is only available for unicast data.
Replay detection
TKIP and CCMP protect against replay attacks by incorporating a sequence counter that is
validated upon receipt. Frames which are "too old" to be valid are discarded.
External protocols and systems
The confidentiality service depends heavily on external protocols to run. Key management is
provided by 802.1X, which together with EAP carries authentication data. 802.11 places no
constraint on the protocols used, but the most common choices are EAP for authentication, and
RADIUS to interface with the authentication server.
Spectrum management services
Spectrum management services are a special subset of station services. They are designed to allow
the wireless network to react to conditions and change radio settings dynamically. Two services were
defined in 802.11h to help meet regulatory requirements.
The first service, transmit power control (TPC), can dynamically adjust the transmission power of a
station. Access points will be able to use the TPC operations to advertise the maximum permissible
power, and reject associations from clients that do not comply with the local radio regulations.
Clients can use TPC to adjust power so that range is "just right" to get to the access point. Digital
cellular systems have a simliar feature designed to extend the battery life of mobile phones.[*] Lower
transmit power also will have some benefit in the form of increased battery life, though the extent of
the improvement will depend on how much the transmit power can be reduced from what the client
would otherwise have used.
[*]
Power control also helps to simplify the electronics in the base station because all signals will be received at roughly the same
signal.
The second service, dynamic frequency selection (DFS), was developed mainly to avoid interfering
with some 5 GHz radar systems in use in Europe. Although originally developed to satisfy European
regulators, the underlying principles have been required by other regulators as well. DFS was key to
the U.S. decision to open up more spectrum in the 5 GHz band in 2004.[ ] DFS includes a way for
the access point to quiet the channel so that it can search for radar without interference, but the
most significant part of DFS is the way that it can reassign the channel on an access point on the fly.
Clients are informed of the new channel just before the channel is switched.
[
]
The decision was made in November 2003, and released as FCC 03-287. The text of the decision is available at
http://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-03-287A1.pdf. Although the spectrum has been allocated, test
procedures were still in development as of this writing, so no FCC-certified devices are yet able to use the new spectrum.
Mobility Support
Mobility is the usually the primary motivation for deploying an 802.11 network. Transmitting data
frames while the station is moving will do for data communications what mobile telephony did for
voice.
802.11 provides mobility between basic service areas at the link layer. However, it is not aware of
anything that happens above the link layer. When designing deploying 802.11, networks engineers
must take care so that the seamless transition at the radio layer is also supported at the network
protocol layer that the station IP address can be preserved. As far as 802.11 is concerned, there are
three types of transitions between access points:
No transition
When stations do not move out of their current access point's service area, no transition is
necessary. This state occurs because the station is not moving or it is moving within the basic
service area of its current access point.[*] (Arguably, this isn't a transition so much as the
absence of a transition, but it is defined in the specification.)
[*]
Although my explanation makes it sound as if the "no motion" and "local motion" substates are easily
distinguishable, they are not. The underlying physics of RF propagation can make it impossible to tell
whether a station is moving because the signal strength can vary with the placement of objects in the
room, which, of course, includes the people who may be walking around.
BSS transition
Stations continuously monitor the signal strength and quality from all access points
administratively assigned to cover an extended service area. Within an extended service area,
802.11 provides MAC layer mobility. Stations attached to the distribution system can send out
frames addressed to the MAC address of a mobile station and let the access points handle the
final hop to the mobile station. Distribution system stations do not need to be aware of a
mobile station's location as long as it is within the same extended service area.
Figure 2-9 illustrates a BSS transition. The three access points in the picture are all assigned to
the same ESS. At the outset, denoted by t=1, the laptop with an 802.11 network card is sitting
within AP1's basic service area and is associated with AP1. When the laptop moves out of AP1's
basic service area and into AP2's at t=2, a BSS transition occurs. The mobile station uses the
reassociation service to associate with AP2, which then starts sending frames to the mobile
station.
BSS transitions require the cooperation of access points. In this scenario, AP2 needs to inform
AP1 that the mobile station is now associated with AP2. 802.11 does not specify the details of
the communications between access points during BSS transitions.
Note that even though two access points are members of the same extended set, they may
nonetheless be connected by a router, which is a layer 3 boundary. In such a scenario, there is
no way to guarantee seamless connectivity using 802.11 protocols only.
Figure 2-9. BSS transition
ESS transition
An ESS transition refers to the movement from one ESS to a second distinct ESS. 802.11 does
not support this type of transition, except to allow the station to associate with an access point
in the second ESS once it leaves the first. Higher-layer connections are almost guaranteed to
be interrupted. It would be fair to say that 802.11 supports ESS transitions only to the extent
that it is relatively easy to attempt associating with an access point in the new extended
service area. Maintaining higher-level connections requires support from the protocol suites in
question. In the case of TCP/ IP, Mobile IP is required to seamlessly support an ESS transition.
Figure 2-10 illustrates an ESS transition. Four basic service areas are organized into two
extended service areas. Seamless transitions from the lefthand ESS to the righthand ESS are
not supported. ESS transitions are supported only because the mobile station will quickly
associate with an access point in the second ESS. Any active network connections are likely to
be dropped when the mobile station leaves the first ESS.
Designing Networks for Mobility
Most networks are designed so that a group of access points provides access to a group of resources.
All the access points under control of the networking organization are assigned to the same SSID,
and clients are configured to use that SSID when connecting to the wireless network.
As client systems move around, they continuously monitor network connectivity, and shift between
access points in the same SSID. 802.11 ensures that clients will be able to move associations
between the access points in the same SSID, but network architects must build the network to
support mobile clients. Small networks are often built on a single VLAN with a single subnet, in which
case there is no need to worry about mobility. Larger networks that span subnet boundaries must
apply some additional technology to provide mobility support. Many products can work
Figure 2-10. ESS transition
with a VLAN core, which allows clients to always attach to the same VLAN throughout an
organization. New products even allow dynamic VLAN assignment based on authentication data.
When users connect to the network, they are attached to the same VLAN everywhere; the switched
network simply requires that the wireless LAN device tag frames appropriately. Some products
support the Mobile IP standard, or use VPN technology creatively. Trade-offs between all the different
mobility strategies are discussed in Chapter 21.
In practice, ESS transitions are quite rare. They usually only occur when users leave one
administrative domain for another (say, the corporate network for a hot spot), in which case the two
networks in question would have different IP addresses and no trust relationship to transparently
attach a client without interrupting network-layer connectivity.
Proprietary mobility systems
Many vendors, especially those who have designed products to build large-scale network
environments, have designed their own protocols and procedures for mobility. When I started
revising this book, one of the big problems with 802.11 was that it generally required a single IP
subnet across an entire roaming area. As a result, wireless LAN deployment often required
substantial architecture work as well as major backbone re-engineering efforts. Several vendors
rushed to fill the gap by implementing proprietary protocols that enabled sessions to move quickly
between access points over arbitrary network topologies. Basic concepts for these solutions will be
discussed in the deployment section of this book.
Chapter 3. 802.11 MAC Fundamentals
This chapter begins our exploration of the 802.11 standard in depth. Chapter 2 provided a high-level
overview of the standard and discussed some of its fundamental attributes. You are now at a fork in
the book. Straight ahead lies a great deal of information on the 802.11 specification and the various
related standards that it uses liberally. It is possible, however, to build a wired network without a
thorough and detailed understanding of the protocols, and the same is true for wireless networks.
However, there are a number of situations in which you may need a deeper knowledge of the
machinery under the hood:
Although 802.11 has been widely and rapidly adopted, security issues have continued to grab
headlines. Network managers will undoubtedly be asked to comment on security issues,
especially in any wireless LAN proposals. To understand and participate in these discussions,
read Chapters 5 and 6. WEP with static keys should be considered fully broken. Solutions based
on 802.1X and dynamic WEP keying are significantly stronger, with the full complement of
protocols in 802.11i described in Chapter 7 stronger still.
Troubleshooting wireless networks is similar to troubleshooting wired networks but can be much
more complex. As always, a trusty packet sniffer can be an invaluable aid. To take full
advantage of a packet sniffer, though, you need to understand what the packets mean to
interpret your network's behavior.
Tuning a wireless network is tied intimately to a number of parameters in the specification, as
well as the behavior of the underlying radio technology. To understand the behavior of your
network and what effect the optimizations will have requires a knowledge of what those
parameters really do and how radio waves travel throughout your environment.
Device drivers may expose low-level knobs and dials for you to play with. Most drivers provide
good defaults for all of the parameters, but some give you freedom to experiment. Open source
software users have the source code and are free to experiment with any and all settings.
Wireless LAN technology is developing rapidly, and new protocol features are constantly being
added. A solid understanding of the base protocol allows you to understand how new features
will function and what they will mean for your network.
As with many other things in life, the more you know, the better off you are. Ethernet is usually
trouble-free, but serious network administrators have long known that when you do run into trouble,
there is no substitute for thorough knowledge of how the network is working. When the first edition
of this book was out, wireless LANs had been given a "free ride." Because they were cool, users were
forgiving when they failed; wireless connectivity was a privilege, not a right. And since there were
relatively few networks and relatively few users on those networks, the networks were rarely
subjected to severe stresses. An Ethernet that has only a half dozen nodes is not likely to be a source
of problems; problems occur when you add a few high-capacity servers, a few hundred users, and
the associated bridges and routers to glue everything together. As the typical 802.11 network grew
up from an access point or two serving a dozen users into a much larger network designed to provide
seamless coverage throughout a building, the stresses on the equipment and protocols has become
much more apparent.
That is why you should read this chapter. Now on to the details. The key to the 802.11 specification
is the MAC. It rides on every physical layer and controls the transmission of user data into the air. It
provides the core framing operations and the interaction with a wired network backbone. Different
physical layers may provide different transmission speeds, all of which are supposed to interoperate.
802.11 does not depart from the previous IEEE 802 standards in any radical way. The standard
successfully adapts Ethernet-style networking to radio links. Like Ethernet, 802.11 uses a carrier
sense multiple access (CSMA) scheme to control access to the transmission medium. However,
collisions waste valuable transmission capacity, so rather than the collision detection (CSMA/CD)
employed by Ethernet, 802.11 uses collision avoidance (CSMA/CA). Also like Ethernet, 802.11 uses a
distributed access scheme with no centralized controller. Each 802.11 station uses the same method
to gain access to the medium. The major differences between 802.11 and Ethernet stem from the
differences in the underlying medium.
This chapter provides some insight into the motivations of the MAC designers by describing some
challenges they needed to overcome and describes the rules used for access to the medium, as well
as the basic frame structure. If you simply want to understand the basic frame sequences that you
will see on an 802.11 network, skip ahead to the end of this chapter. For further information on the
MAC, consult its formal specification in Clause 9 of the 802.11 standard; detailed MAC state diagrams
are in Annex C.
Challenges for the MAC
Differences between the wireless network environment and the traditional wired environment create
challenges for network protocol designers. This section examines a number of the hurdles that the
802.11 designers faced.
RF Link Quality
On a wired Ethernet, it is reasonable to transmit a frame and assume that the destination receives it
correctly. Radio links are different, especially when the frequencies used are unlicensed ISM bands.
Even narrowband transmissions are subject to noise and interference, but unlicensed devices must
assume that interference will exist and work around it. The designers of 802.11 considered ways to
work around the radiation from microwave ovens and other RF sources. In addition to the noise,
multipath fading may also lead to situations in which frames cannot be transmitted because a node
moves into a dead spot.
Unlike many other link layer protocols, 802.11 incorporates positive acknowledgments. All
transmitted frames must be acknowledged, as shown in Figure 3-1. If any part of the transfer fails,
the frame is considered lost.
Figure 3-1. Positive acknowledgment of data transmissions
The sequence in Figure 3-1 is an atomic operation, which means it is a single transactional unit.
Although there are multiple steps in the transaction, it is considered a single indivisible operation.
Atomic operations are "all or nothing." Either every step in the sequence must complete successfully,
or the entire operation is considered a failure. The sender of the data frame must receive an
acknowledgment, or the frame is considered lost. It does not matter from the sender's perspective
whether the initial data frame was lost in transit, or the corresponding acknowledgment was lost in
transit. In either case, the data frame must be retransmitted.
One of the additional complexities of treating the frame transmission of Figure 3-1 as atomic is that
the transaction occurs in two pieces, subject to control by two different stations. Both stations must
work together to jointly take control of the network medium for transmissions during the entire
transaction. 802.11 allows stations to lock out contention during atomic operations so that atomic
sequences are not interrupted by other stations attempting to use the transmission medium.
Radio link quality also influences the speed at which a network can operate. Good quality signals can
carry data at a higher speed. Signal quality degrades with range, which means that the data
transmission speed of an 802.11 station depends on its location relative to the access point. Stations
must implement a method for determining when to change the data rate in response to changing
conditions. Furthermore, the complete collection of stations in a network must manage transmissions
at multiple speeds. Rules for multirate support are discussed later in this chapter.
The Hidden Node Problem
In Ethernet networks, stations depend on the reception of transmissions to perform the carrier
sensing functions of CSMA/CD. Wires in the physical medium contain the signals and distribute them
to network nodes. Wireless networks have fuzzier boundaries, sometimes to the point where each
node may not be able to directly communicate with every other node in the wireless network, as in
Figure 3-2.
Figure 3-2. Nodes 1 and 3 are "hidden"
In the Figure 3-2, node 2 can communicate with both nodes 1 and 3, but something prevents nodes
1 and 3 from communicating directly. (The obstacle itself is not relevant; it could be as simple as
nodes 1 and 3 being as far away from 2 as possible, so the radio waves cannot reach the full distance
from 1 to 3.) From the perspective of node 1, node 3 is a "hidden" node. If a simple transmit-andpray protocol was used, it would be easy for node 1 and node 3 to transmit simultaneously, thus
rendering node 2 unable to make sense of anything. Furthermore, nodes 1 and 3 would not have any
indication of the error because the collision was local to node 2.
Collisions resulting from hidden nodes may be hard to detect in wireless networks because wireless
transceivers are generally half-duplex; they don't transmit and receive at the same time. To prevent
collisions, 802.11 allows stations to use Request to Send (RTS) and Clear to Send (CTS) signals to
clear out an area. Both the RTS and CTS frames extend the frame transaction, so that the RTS
frame, CTS frame, the data frame, and the final acknowledgment are all considered part of the same
atomic operation. Figure 3-3 illustrates the procedure.
Figure 3-3. RTS/CTS clearing
In Figure 3-3, node 1 has a frame to send; it initiates the process by sending an RTS frame. The RTS
frame serves several purposes: in addition to reserving the radio link for transmission, it silences any
stations that hear it. If the target station receives an RTS, it responds with a CTS. Like the RTS
frame, the CTS frame silences stations in the immediate vicinity. Once the RTS/CTS exchange is
complete, node 1 can transmit its frames without worry of interference from any hidden nodes.
Hidden nodes beyond the range of the sending station are silenced by the CTS from the receiver.
When the RTS/CTS clearing procedure is used, any frames must be positively acknowledged.
The multiframe RTS/CTS transmission procedure consumes a fair amount of capacity, especially
because of the additional latency incurred before transmission can commence. As a result, it is used
only in high-capacity environments and environments with significant contention on transmission. For
lower-capacity environments, it is not necessary.
Hidden nodes have also become less of a problem as 802.11 has grown up. In small, quiescent
networks with only a few stations associated to an access point, there is very little risk of
simultaneous transmission, and plenty of spare capacity to be used for retransmission. In many
larger environments, the coverage is dense enough that the clients are located physically close
enough to an access point that they are all within range of each other. (In fact, the range of many
client systems is probably too large for most networks, which will be explored in the planning phase
of this book.)
You can control the RTS/CTS procedure by setting the RTS threshold if the device driver for your
802.11 card allows you to adjust it. The RTS/CTS exchange is performed for frames larger than the
threshold. Frames shorter than the threshold are simply sent.
MAC Access Modes and Timing
Access to the wireless medium is controlled by coordination functions. Ethernet-like CSMA/CA access
is provided by the distributed coordination function (DCF). If contention-free service is required, it
can be provided by the point coordination function (PCF), which is built on top of the DCF. Between
the free-for-all of the DCF and the precision of the PCF, networks can use the hybrid coordination
function (HCF), a middle ground for quality of service between the two extremes. Contention-free
services are provided only in infrastructure networks, but quality of service may be provided in any
network that has HCF support in the stations. The coordination functions are described in the
following list and illustrated in Figure 3-4:
DCF
The DCF is the basis of the standard CSMA/CA access mechanism. Like Ethernet, it first checks
to see that the radio link is clear before transmitting. To avoid collisions, stations use a random
backoff after each frame, with the first transmitter seizing the channel. In some circumstances,
the DCF may use the CTS/RTS clearing technique to further reduce the possibility of collisions.
PCF
The point coordination function provides contention-free services. Special stations called point
coordinators are used to ensure that the medium is provided without contention. Point
coordinators reside in access points, so the PCF is restricted to infrastructure networks. To gain
priority over standard contention-based services, the PCF allows stations to transmit frames
after a shorter interval. The PCF is not widely implemented and is described in Chapter 9.
HCF
Some applications need to have service quality that is a step above best-effort delivery, but the
rigorous timing of the PCF is not required. The HCF allows stations to maintain multiple service
queues and balance access to the wireless medium in favor of applications that require better
service quality. The HCF is not fully standardized yet, but is being produced as part of the
eventual 802.11e specification.
Adding quality of service to the 802.11 MAC was a significant undertaking. Due to the added
complexity in terms of framing, queue management, and signaling, the quality of service
specifications were still subject to standards committee wrangling as this book was written, and
discussion of them must be postponed to a future edition.
Carrier-Sensing Functions and the Network Allocation Vector
Carrier sensing is used to determine if the medium is available. Two types of carrier-sensing
functions in 802.11 manage this process: the physical carrier-sensing and virtual carrier-sensing
functions. If either carrier-sensing function indicates that the medium is busy, the MAC reports this to
higher layers.
Figure 3-4. MAC coordination functions
Physical carrier-sensing functions are provided by the physical layer in question and depend on the
medium and modulation used. It is difficult (or, more to the point, expensive) to build physical
carrier-sensing hardware for RF-based media, because transceivers can transmit and receive
simultaneously only if they incorporate expensive electronics. Furthermore, with hidden nodes
potentially lurking everywhere, physical carrier-sensing cannot provide all the necessary information.
Virtual carrier-sensing is provided by the Network Allocation Vector (NAV). Most 802.11 frames carry
a duration field, which can be used to reserve the medium for a fixed time period. The NAV is a timer
that indicates the amount of time the medium will be reserved, in microseconds. Stations set the
NAV to the time for which they expect to use the medium, including any frames necessary to
complete the current operation. Other stations count down from the NAV to 0. When the NAV is
nonzero, the virtual carrier-sensing function indicates that the medium is busy; when the NAV
reaches 0, the virtual carrier-sensing function indicates that the medium is idle.
By using the NAV, stations can ensure that atomic operations are not interrupted. For example, the
RTS/CTS sequence in Figure 3-3 is atomic. Figure 3-5 shows how the NAV protects the sequence
from interruption. (This is a standard format for a number of diagrams in this book that illustrate the
interaction of multiple stations with the corresponding timers.) Activity on the medium by stations is
represented by the shaded bars, and each bar is labeled with the frame type. Interframe spacing is
depicted by the lack of any activity. Finally, the NAV timer is represented by the bars on the NAV line
at the bottom of the figure. The NAV is carried in the frame headers on the RTS and CTS frames; it is
depicted on its own line to show how the NAV relates to actual transmissions in the air. When a NAV
bar is present on the NAV line, stations should defer access to the medium because the virtual
carrier-sensing mechanism will indicate a busy medium.
To ensure that the sequence is not interrupted, node 1 sets the NAV in its RTS to block access to the
medium while the RTS is being transmitted. All stations that hear the RTS defer access to the
medium until the NAV elapses.
Figure 3-5. Using the NAV for virtual carrier sensing
RTS frames are not necessarily heard by every station in the network. Therefore, the recipient of the
intended transmission responds with a CTS that includes a shorter NAV. This NAV prevents other
stations from accessing the medium until the transmission completes. After the sequence completes,
the medium can be used by any station after distributed interframe space (DIFS), which is depicted
by the contention window beginning at the right side of the figure.
RTS/CTS exchanges may be useful in crowded areas with multiple overlapping networks. Every
station on the same physical channel receives the NAV and defers access appropriately, even if the
stations are configured to be on different networks.
Interframe Spacing
As with traditional Ethernet, the interframe spacing plays a large role in coordinating access to the
transmission medium. 802.11 uses four different interframe spaces. Three are used to determine
medium access; the relationship between them is shown in Figure 3-6.
Figure 3-6. Interframe spacing relationships
We've already seen that as part of the collision avoidance built into the 802.11 MAC, stations delay
transmission until the medium becomes idle. Varying interframe spacings create different priority
levels for different types of traffic. The logic behind this is simple: high-priority traffic doesn't have to
wait as long after the medium has become idle. Therefore, if there is any high-priority traffic waiting,
it grabs the network before low-priority frames have a chance to try. To assist with interoperability
between different data rates, the interframe space is a fixed amount of time, independent of the
transmission speed. (This is only one of the many problems caused by having different physical
layers use the same radio resources, which are different modulation techniques.) Different physical
layers, however, can specify different interframe space times.
Short interframe space (SIFS)
The SIFS is used for the highest-priority transmissions, such as RTS/CTS frames and positive
acknowledgments. High-priority transmissions can begin once the SIFS has elapsed. Once
these high-priority transmissions begin, the medium becomes busy, so frames transmitted
after the SIFS has elapsed have priority over frames that can be transmitted only after longer
intervals.
PCF interframe space (PIFS)
The PIFS, sometimes erroneously called the priority interframe space, is used by the PCF
during contention-free operation. Stations with data to transmit in the contention-free period
can transmit after the PIFS has elapsed and preempt any contention-based traffic.
DCF interframe space (DIFS)
The DIFS is the minimum medium idle time for contention-based services. Stations may have
immediate access to the medium if it has been free for a period longer than the DIFS.
Extended interframe space (EIFS)
The EIFS is not illustrated in Figure 3-6 because it is not a fixed interval. It is used only when
there is an error in frame transmission.
Interframe spacing and priority
Atomic operations start like regular transmissions: they must wait for the the relevant interframe
space, typically the DIFS, to end before they can begin. However, the second and any subsequent
steps in an atomic operation take place after the SIFS, rather than after the DIFS. Because the SIFS
is shorter than the other interframe spaces, the second (and subsequent) parts of an atomic
operation will grab the medium before another type of frame can be transmitted. By using the SIFS
and the NAV, stations can seize the medium for as long as necessary.
In Figure 3-5, for example, the short interframe space is used between the different units of the
atomic exchange. After the sender gains access to the medium, the receiver replies with a CTS after
the SIFS. Any stations that might attempt to access the medium at the conclusion of the RTS would
wait for one DIFS interval. Partway through the DIFS interval, though, the SIFS interval elapses, and
the CTS is transmitted.
Contention-Based Access Using the DCF
Most traffic uses the DCF, which provides a standard Ethernet-like contention-based service. The DCF
allows multiple independent stations to interact without central control, and thus may be used in
either IBSS networks or in infrastructure networks.
Before attempting to transmit, each station checks whether the medium is idle. If the medium is not
idle, stations defer to each other and employ an orderly exponential backoff algorithm to avoid
collisions.
In distilling the 802.11 MAC rules, there is a basic set of rules that are always used, and additional
rules may be applied depending on the circumstances. Two basic rules apply to all transmissions
using the DCF:
1. If the medium has been idle for longer than the DIFS, transmission can begin immediately.
Carrier sensing is performed using both a physical medium-dependent method and the virtual
(NAV) method.
a. If the previous frame was received without errors, the medium must be free for at least
the DIFS.
b. If the previous transmission contained errors, the medium must be free for the amount of
the EIFS.
2. If the medium is busy, the station must wait for the channel to become idle. 802.11 refers to
the wait as access deferral. If access is deferred, the station waits for the medium to be idle for
the DIFS and prepares for the exponential backoff procedure.
Additional rules may apply in certain situations. Many of these rules depend on the particular
situation "on the wire" and are specific to the results of previous transmissions.
1. Error recovery is the responsibility of the station sending a frame. Senders expect
acknowledgments for each transmitted frame and are responsible for retrying the transmission
until it is successful.
a. Positive acknowledgments are the only indication of success. Atomic exchanges must
complete in their entirety to be successful. If an acknowledgment is expected and does not
arrive, the sender considers the transmission lost and must retry.
b. All unicast data must be acknowledged. Broadcast data is not acknowledged. (As a result,
unicast data has an inherently higher service quality than broadcast data, even though the
radio link is inherently a broadcast medium.)
c. Any failure increments a retry counter, and the transmission is retried. A failure can be
2.
c.
due to a failure to gain access to the medium or a lack of an acknowledgment. However,
there is a longer congestion window when transmissions are retried (see next section).
2. Multiframe sequences may update the NAV with each step in the transmission procedure. When
a station receives a medium reservation that is longer than the current NAV, it updates the
NAV. Setting the NAV is done on a frame-by-frame basis and is discussed in much more detail
in the next chapter.
3. The following types of frames can be transmitted after the SIFS and thus receive maximum
priority: acknowledgments, the CTS in an RTS/CTS exchange sequence, and fragments in
fragment sequences.
a. Once a station has transmitted the first frame in a sequence, it has gained control of the
channel. Any additional frames and their acknowledgments can be sent using the short
interframe space, which locks out any other stations.
b. Additional frames in the sequence update the NAV for the expected additional time the
medium will be used.
4. Extended frame sequences are required for higher-level packets that are larger than configured
thresholds.
a. Packets larger than the RTS threshold must have RTS/CTS exchange.
b. Packets larger than the fragmentation threshold must be fragmented.
Error Recovery with the DCF
Error detection and correction is up to the station that begins an atomic frame exchange. When an
error is detected, the station with data must resend the frame. Errors must be detected by the
sending station. In some cases, the sender can infer frame loss by the lack of a positive
acknowledgment from the receiver. Retry counters are incremented when frames are retransmitted.
Each frame or fragment has a single retry counter associated with it. Stations have two retry
counters: the short retry count and the long retry count. Frames that are shorter than the RTS
threshold are considered to be short; frames longer than the threshold are long. Depending on the
length of the frame, it is associated with either a short or long retry counter. Frame retry counts
begin at 0 and are incremented when a frame transmission fails.
The short retry count is reset to 0 when:
A CTS frame is received in response to a transmitted RTS
A MAC-layer acknowledgment is received after a nonfragmented transmission
A broadcast or multicast frame is received
The long retry count is reset to 0 when:
A MAC-layer acknowledgment is received for a frame longer than the RTS threshold
A broadcast or multicast frame is received
In addition to the associated retry count, fragments are given a maximum "lifetime" by the MAC.
When the first fragment is transmitted, the lifetime counter is started. When the lifetime limit is
reached, the frame is discarded and no attempt is made to transmit any remaining fragments.
Naturally, higher-layer protocols may detect any loss and retransmit the data; when a higher-level
protocol such as TCP retransmits data, though, it would be a new frame to the 802.11 MAC and all
the retry counters will be restarted.
Using the retry counters
Like most other network protocols, 802.11 provides reliability through retransmission. Data
transmission happens within the confines of an atomic sequence, and the entire sequence must
complete for a transmission to be successful. When a station transmits a frame, it must receive an
acknowledgment from the receiver or it will consider the transmission to have failed. Failed
transmissions increment the retry counter associated with the frame (or fragment). If the retry limit
is reached, the frame is discarded, and its loss is reported to higher-layer protocols.
One of the reasons for having short frames and long frames is to allow network administrators to
customize the robustness of the network for different frame lengths. Large frames require more
buffer space, so one potential application of having two separate retry limits is to decrease the long
retry limit to decrease the amount of buffer space required.
Backoff with the DCF
After frame transmission has completed and the DIFS has elapsed, stations may attempt to transmit
congestion-based data. A period called the contention window or backoff window follows the DIFS.
This window is divided into slots. Slot length is medium-dependent; higher-speed physical layers use
shorter slot times. Stations pick a random slot and wait for that slot before attempting to access the
medium; all slots are equally likely selections. When several stations are attempting to transmit, the
station that picks the first slot (the station with the lowest random number) wins. According to the
standard, all slot numbers should be equally likely; see the sidebar on Spectralink Voice Priority later
in this chapter for one notable exception.
As in Ethernet, the backoff time is selected from a larger range each time a transmission fails. Figure
3-7 illustrates the growth of the contention window as the number of transmissions increases, using
the numbers from the 802.11b direct-sequence spread-spectrum (DSSS) physical layer. Other
physical layers use different sizes, but the principle is identical. Contention window sizes are always 1
less than a power of 2 (e.g., 31, 63, 127, 255). Each time the retry counter increases, the contention
window moves to the next greatest power of two. The size of the contention window is limited by the
physical layer. For example, the DS physical layer limits the contention window to 1,023 transmission
slots.
When the contention window reaches its maximum size, it remains there until it can be reset.
Allowing long contention windows when several competing stations are attempting to gain access to
the medium keeps the MAC algorithms stable even under maximum load. The contention window is
reset to its minimum size when frames are transmitted successfully, or the associated retry counter
is reached, and the frame is discarded.
Figure 3-7. DSSS contention window size
Fragmentation and Reassembly
Higher-level packets and some large management frames may need to be broken into smaller pieces
to fit through the wireless channel. Fragmentation may also help improve reliability in the presence
of interference. Wireless LAN stations may attempt to fragment transmissions so that interference
affects only small fragments, not large frames. By immediately reducing the amount of data that can
be corrupted by interference, fragmentation may result in a higher effective throughput. Interference
may come from a variety of sources. Some, but by no means all, microwave ovens cause
interference with 2.4 GHz networks.[*] Electromagnetic radiation is generated by the magnetron tube
during its ramp-up and ramp-down, so microwaves emit interference half the time. Many newer
cordless phones also cause interference.[ ] Outdoor networks are subject to a much wider variety of
interference.
[*]
In the U.S., appliances are powered by 60-Hz alternating current, so microwaves interfere for about 8 milliseconds (ms) out of
every 16-ms cycle. Much of the rest of the world uses 50-Hz current, and interference takes place for 10 ms out of the 20-ms
cycle.
[
] If you need to use a cordless phone in the same area as a wireless LAN, I suggest purchasing a 900 MHz cordless phone on
eBay.
Wireless LAN stations may attempt to fragment transmissions so that interference affects only small
fragments, not large frames. By immediately reducing the amount
Spectralink Voice Priority
One of the challenges in supporting voice on wireless networks is that voice is far more
sensitive to poor network service than data applications. If a 1,500 byte fragment of a
graphics file is a tenth of a second late, the typical user will not even notice. If a delay of
a tenth of a second is introduced into a phone conversation, though, it will be too much.
Providing high-quality service over an IP network is hard enough. Doing so over a
wireless LAN is doubly challenging. One of the major problems that network engineers
face in designing wireless LAN voice networks is that all data is treated equally. If there
is a short voice frame and a long data frame, there is no inherent preference for one or
the other.
Spectralink, a manufacturer of handheld 802.11 phones, has devised a special set of
protocol extensions, called Spectralink Voice Priority (SVP), to assist in making the
network more useful for voice transport. SVP consists of components implemented in
both access points and in handsets to prioritze voice over data and coordinate several
voice calls on a single AP. SVP assists with both the downlink from the AP and the uplink
from handsets.
To support SVP, an access point must transmit voice frames with zero backoff. Rather
than selecting a backoff slot number as required by the 802.11 standard, access points
with SVP enabled will always choose zero. In the presence of contention for the wireless
medium, the voice frames with zero backoff will have de facto priority boost because
data frames are likely to have a positive backoff slot. Strictly speaking, stations
implementing zero backoff are no longer compliant with 802.11 because it mandates
selection of a backoff slot in accordance with defined rules. (To preserve stability under
load, however, retransmitted voice frames are subject to the backoff rules.)
By selecting zero backoff, access points implementing SVP ensure that voice frames have
preferential access to the air. Access points that implement SVP must also keep track of
voice frames and provide preferential queuing treatment as well. SVP requires that voice
frames be pushed to the head of the queue for transmission. APs implement transmit
queues in many different ways; the important point is the functional result, which is that
voice frames move to the head of the line. Some APs may move voice frames up to the
head of a single transmit queue, while other APs may maintain multiple transmit queues
and serve the high-priority voice queue first.
of data that can be corrupted by interference, fragmentation may result in a higher effective
throughput. Fragmentation takes place when the length of a higher-level packet exceeds the
fragmentation threshold configured by the network administrator. Fragments all have the same
frame sequence number but have ascending fragment numbers to aid in reassembly. Frame control
information also indicates whether more fragments are coming. All of the fragments that comprise a
frame are normally sent in a fragmentation burst, which is shown in Figure 3-8. This figure also
incorporates an RTS/CTS exchange, because it is common for the fragmentation and RTS/CTS
thresholds to be set to the same value. The figure also shows how the NAV and SIFS are used in
combination to control access to the medium.
Figure 3-8. Fragmentation burst
Fragments and their acknowledgments are separated by the SIFS, so a station retains control of the
channel during a fragmentation burst. The NAV is also used to ensure that other stations do not use
the channel during the fragmentation burst. As with any RTS/CTS exchange, the RTS and CTS both
set the NAV from the expected time to the end of the first fragments in the air. Subsequent
fragments then form a chain. Each fragment sets the NAV to hold the medium until the end of the
acknowledgment for the next frame. Fragment 0 sets the NAV to hold the medium until ACK 1,
fragment 1 sets the NAV to hold the medium until ACK 2, and so on. After the last fragment and its
acknowledgment have been sent, the NAV is set to 0, indicating that the medium will be released
after the fragmentation burst completes.
Frame Format
To meet the challenges posed by a wireless data link, the MAC was forced to adopt several unique
features, not the least of which was the use of four address fields. Not all frames use all the address
fields, and the values assigned to the address fields may change depending on the type of MAC
frame being transmitted. Details on the use of address fields in different frame types are presented
in Chapter 4. Figure 3-9 shows the generic 802.11 MAC frame. All diagrams in this section follow the
IEEE conventions in 802.11. Fields are transmitted from left to right.
Figure 3-9. Generic 802.11 MAC frame
802.11 MAC frames do not include some of the classic Ethernet frame features, most notably the
type/length field and the preamble. The preamble is part of the physical layer, and encapsulation
details such as type and length are present in the header on the data carried in the 802.11 frame.
Frame Control
Each frame starts with a two-byte Frame Control subfield, shown in Figure 3-10. The components of
the Frame Control subfield are:
Protocol version
Two bits indicate which version of the 802.11 MAC is contained in the rest of the frame. At
present, only one version of the 802.11 MAC has been developed; it is assigned the protocol
number 0. Other values will appear when the IEEE standardizes changes to the MAC that
render it incompatible with the initial specification. So far, none of the revisions to 802.11 have
required incrementing the protocol number.
Type and subtype fields
Type and subtype fields identify the type of frame used. To cope with noise and unreliability, a
number of management functions are incorporated into the 802.11 MAC. Some, such as the
RTS/CTS operations and the acknowledgments, have already been discussed. Table 3-1 shows
how the type and subtype identifiers are used to create the different classes of frames.
Figure 3-10. Frame control field
In Table 3-1, bit strings are written most-significant bit first, which is the reverse of the order used in
Figure 3-10. Therefore, the frame type is the third bit in the frame control field followed by the
second bit (b3 b2), and the subtype is the seventh bit, followed by the sixth, fifth, and fourth bits (b7
b6 b5 b4).
Table 3-1. Type and subtype identifiers
Subtype value
Subtype name
Management frames (type=00)a
0000
Association request
0001
Association response
0010
Reassociation request
0011
Reassociation response
0100
Probe request
0101
Probe response
1000
Beacon
1001
Announcement traffic indication message (ATIM)
1010
Disassociation
1011
Authentication
1100
Deauthentication
1101
Action (for spectrum management with 802.11h, also for QoS)
Control frames (type=01)b
1000
Block Acknowledgment Request (QoS)
1001
Block Acknowledgment (QoS)
1010
Power Save (PS)-Poll
1011
RTS
1100
CTS
1101
Acknowledgment (ACK)
Subtype value
Subtype name
1110
Contention-Free (CF)-End
1111
CF-End+CF-Ack
Data frames (type=10)
0000
Data
0001
Data+CF-Ack
0010
Data+CF-Poll
0011
Data+CF-Ack+CF-Poll
0100
Null data (no data transmitted)
0101
CF-Ack (no data transmitted)
0110
CF-Poll (no data transmitted)
0111
CF-Ack+CF-Poll (no data transmitted)
1000
QoS Datac
1001
QoS Data + CF-Ackc
1010
QoS Data + CF-Pollc
1011
QoS Data + CF-Ack + CF-Pollc
1100
QoS Null (no data transmitted)c
1101
QoS CF-Ack (no data transmitted)c
1110
QoS CF-Poll (no data transmitted)c
1111
QoS CF-Ack+CF-Poll (no data transmitted)c
(Frame type 11 is reserved)
a
Management subtypes 0110-0111 and 1110-1111 are reserved and not currently used.
b
Control subtypes 0000-0111 are reserved and not currently used.
c
Proposed by the 802.11e task group, but not yet standardized. Note that these frames all have a
leading one, which has caused some to refer to the first bit as the QoS bit.
ToDS and FromDS bits
These bits indicate whether a frame is destined for the distribution system. All frames on
infrastructure networks will have one of the distribution system bits set. Table 3-2 shows how
these bits are interpreted. As Chapter 4 will explain, the interpretation of the address fields
depends on the setting of these bits.
Table 3-2. Interpreting the ToDS and FromDS bits
To DS=0
To DS=1
From
DS=0
All management and control frames Data
frames within an IBSS (never infrastructure
data frames)
Data frames transmitted from a
wireless station in an infrastructure
network.
From
DS=1
Data frames received for a wireless station in
an infrastructure network
Data frames on a "wireless bridge"
More fragments bit
This bit functions much like the "more fragments" bit in IP. When a higher-level packet has
been fragmented by the MAC, the initial fragment and any following nonfinal fragments set this
bit to 1. Large data frames and some management frames may be large enough to require
fragmentation; all other frames set this bit to 0. In practice, most data frames are transmitted
at the maximum Ethernet size and fragmentation is not often used.
Retry bit
From time to time, frames may be retransmitted. Any retransmitted frames set this bit to 1 to
aid the receiving station in eliminating duplicate frames.
Power management bit
Network adapters built on 802.11 are often built to the PC Card form factor and used in
battery-powered laptop or handheld computers. To conserve battery life, many small devices
have the ability to power down parts of the network interface. This bit indicates whether the
sender will be in a powersaving mode after the completion of the current atomic frame
exchange. 1 indicates that the station will be in powersave mode, and 0 indicates that the
station will be active. Access points perform a number of important management functions and
are not allowed to save power, so this bit is always 0 in frames transmitted by an access point.
More data bit
To accommodate stations in a powersaving mode, access points may buffer frames received
from the distribution system. An access point sets this bit to indicate that at least one frame is
available and is addressed to a dozing station.
Protected Frame bit
Wireless transmissions are inherently easier to intercept than transmissions on a fixed network.
If the frame is protected by link layer security protocols, this bit is set to 1, and the frame
changes slightly. The Protected Frame bit was previously called the WEP bit.
Order bit
Frames and fragments can be transmitted in order at the cost of additional processing by both
the sending and receiving MACs. When the "strict ordering" delivery is employed, this bit is set
to 1.
Duration/ID Field
The Duration/ID field follows the frame control field. This field has several uses and takes one of the
three forms shown in Figure 3-11.
Figure 3-11. Duration/ID field
Duration: setting the NAV
When bit 15 is 0, the duration/ID field is used to set the NAV. The value represents the number of
microseconds that the medium is expected to remain busy for the transmission currently in progress.
All stations must monitor the headers of all frames they receive and update the NAV accordingly. Any
value that extends the amount of time the medium is busy updates the NAV and blocks access to the
medium for additional time.
Frames transmitted during contention-free periods
During the contention-free periods, bit 14 is 0 and bit 15 is 1. All other bits are 0, so the duration/ID
field takes a value of 32,768. This value is interpreted as a NAV. It allows any stations that did not
receive the Beacon[*] announcing the contention-free period to update the NAV with a suitably large
value to avoid interfering with contention-free transmissions.
[*]
Beacon frames are a subtype of management frames, which is why "Beacon" is capitalized.
PS-Poll frames
Bits 14 and 15 are both set to 1 in PS-Poll frames. Mobile stations may elect to save battery power
by turning off antennas. Dozing stations must wake up periodically. To ensure that no frames are
lost, stations awaking from their slumber transmit a PS-Poll frame to retrieve any buffered frames
from the access point. Along with this request, waking stations incorporate the association ID (AID)
that indicates which BSS they belong to. The AID is included in the PS-Poll frame and may range
from 1-2,007. Values from 2,008-16,383 are reserved and not used.
Address Fields
An 802.11 frame may contain up to four address fields. The address fields are numbered because
different fields are used for different purposes depending on the frame type (details are found in
Chapter 4). The general rule of thumb is that Address 1 is used for the receiver, Address 2 for the
transmitter, and Address 3 field for filtering by the receiver. In an infrastructure network, for
example, the third address field is used by the receiver to determine whether the frame is part of the
network it is associated to.[*]
[*]
802.11 specifies that stations should ignore frames that do not have the same BSSID, but most products do not correctly
implement BSSID filtering and will pass any received frame up to higher protocol layers.
Addressing in 802.11 follows the conventions used for the other IEEE 802 networks, including
Ethernet. Addresses are 48 bits long. If the first bit sent to the physical medium is a 0, the address
represents a single station (unicast). When the first bit is a 1, the address represents a group of
physical stations and is called a multicast (or group) address. If all bits are 1s, then the frame is a
broadcast and is delivered to all stations connected to the wireless medium.
48-bit addresses are used for a variety of purposes:
Destination address
As in Ethernet, the destination address is the 48-bit IEEE MAC identifier that corresponds to the
final recipient: the station that will hand the frame to higher protocol layers for processing.
Source address
This is the 48-bit IEEE MAC identifier that identifies the source of the transmission. Only one
station can be the source of a frame, so the Individual/Group bit is always 0 to indicate an
individual station.
Receiver address
This is a 48-bit IEEE MAC identifier that indicates which wireless station should process the
frame. If it is a wireless station, the receiver address is the destination address. For frames
destined to a node on an Ethernet connected to an access point, the receiver is the wireless
interface in the access point, and the destination address may be a router attached to the
Ethernet.
Transmitter address
This is a 48-bit IEEE MAC address to identify the wireless interface that transmitted the frame
onto the wireless medium. The transmitter address is used only in wireless bridging.
Basic Service Set ID (BSSID)
To identify different wireless LANs in the same area, stations may be assigned to a BSS. In
infrastructure networks, the BSSID is the MAC address used by the wireless interface in the
access point. Ad hoc networks generate a random BSSID with the Universal/Local bit set to 1
to prevent conflicts with officially assigned MAC addresses.
The number of address fields used depends on the type of frame. Most data frames use three fields
for source, destination, and BSSID. The number and arrangement of address fields in a data frame
depends on how the frame is traveling relative to the distribution system. Most transmissions use
three addresses, which is why only three of the four addresses are contiguous in the frame format.
Sequence Control Field
This 16-bit field is used for both defragmentation and discarding duplicate frames. It is composed of
a 4-bit fragment number field and a 12-bit sequence number field, as shown in Figure 3-12.
Sequence numbers are not used in control frames, so the Sequence Control field is not present.
Figure 3-12. Sequence Control field
Higher-level frames are each given a sequence number as they are passed to the MAC for
transmission. The sequence number subfield operates as a modulo-4096 counter of the frames
transmitted. It begins at 0 and increments by 1 for each higher-level packet handled by the MAC. If
higher-level packets are fragmented, all fragments will have the same sequence number. When
frames are retransmitted, the sequence number is not changed.
What differs between fragments is the fragment number. The first fragment is given a fragment
number of 0. Each successive fragment increments the fragment number by 1. Retransmitted
fragments keep their original sequence numbers to assist in reassembly.
Stations that implement the QoS extensions use a slightly different interpretation of the sequence
control field because multiple transmit queues need to be maintained.
Frame Body
The frame body, also called the Data field, moves the higher-layer payload from station to station. As
originally specified, 802.11 can transmit frames with a maximum payload of 2,304 bytes of higher-
level data. Implementations must support larger frame bodies to accommodate additional headers
for security and QoS. 802.2 LLC headers use 8 bytes for a maximum network protocol payload of
2,296 bytes. Preventing fragmentation must be done at the protocol layer. On IP networks, Path MTU
Discovery (RFC 1191) will prevent the transmission of frames with Data fields larger than 1,500
bytes.
802.11 differs from other link layer technologies in two notable ways. First, there is no higher-level
protocol tag in the 802.11 frame to distinguish between higher-layer protocol types. Higher-level
protocols are tagged with a type field by an additional header, which is used as the start of the
802.11 payload. Second, 802.11 does not generally pad frames to a minimum length. Many frames
used by 802.11 are short, and the chips and electronics used in network interfaces has progressed to
the point where a pad is no longer necessary.
Frame Check Sequence
As with Ethernet, the 802.11 frame closes with a frame check sequence (FCS). The FCS is often
referred to as the cyclic redundancy check (CRC) because of the underlying mathematical operations.
The FCS allows stations to check the integrity of received frames. All fields in the MAC header and
the body of the frame are included in the FCS. Although 802.3 and 802.11 use the same method to
calculate the FCS, the MAC header used in 802.11 is different from the header used in 802.3, so the
FCS must be recalculated by access points.
When frames are sent to the wireless interface, the FCS is calculated before those frames are sent
out over the wireless link. Receivers can then calculate the FCS from the received frame and
compare it to the received FCS. If the two match, there is a high probability that the frame was not
damaged in transit.
On Ethernets, frames with a bad FCS are simply discarded, and frames with a good FCS are passed
up the protocol stack. On 802.11 networks, frames that pass the integrity check may also require the
receiver to send an acknowledgment. For example, data frames that are received correctly must be
positively acknowledged, or they are retransmitted. 802.11 does not have a negative
acknowledgment for frames that fail the FCS; stations must wait for the acknowledgment timeout
before retransmitting.
Encapsulation of Higher-Layer Protocols Within 802.11
Like all other 802 link layers, 802.11 can transport any network-layer protocol. Unlike Ethernet,
802.11 relies on 802.2 logical-link control (LLC) encapsulation to carry higher-level protocols. Figure
3-13 shows how 802.2 LLC encapsulation is used to carry an IP packet. In the figure, the "MAC
headers" for 802.1H and RFC 1042 might be the 12 bytes of source and destination MAC address
information on Ethernet or the long 802.11 MAC header from the previous section.
Figure 3-13. IP encapsulation in 802.11
Two different methods can be used to encapsulate LLC data for transmission. One is described in RFC
1042, and the other in 802.1H. Both standards may go by other names. RFC 1042 is sometimes
referred to as IETF encapsulation, while 802.1H is sometimes called tunnel encapsulation.
As you can see in Figure 3-13, though, the two methods are quite similar. An Ethernet frame is
shown in the top line of Figure 3-13. It has a MAC header composed of source and destination MAC
addresses, a type code, the embedded packet, and a frame check field. In the IP world, the Type
code is either 0x0800 (2048 decimal) for IP itself, or 0x0806 (2054 decimal) for the Address
Resolution Protocol (ARP).
Both RFC 1042 and 802.1H are derivatives of 802.2's sub-network access protocol (SNAP). The MAC
addresses are copied into the beginning of the encapsulation frame, and then a SNAP header is
inserted. SNAP headers begin with a destination service access point (DSAP) and a source service
access point (SSAP). After the addresses, SNAP includes a Control header. Like high-level data link
control (HDLC) and its progeny, the Control field is set to 0x03 to denote unnumbered information
(UI), a category that maps well to the best-effort delivery of IP datagrams. The last field inserted by
SNAP is an organizationally unique identifier (OUI). Initially, the IEEE hoped that the 1-byte service
access points would be adequate to handle the number of network protocols, but this proved to be an
overly optimistic assessment of the state of the world. As a result, SNAP copies the type code from
the original Ethernet frame. The only difference between 802.1H and RFC 1042 is the OUI used.
At one point, many products offered the option to switch between the two encapsulation standards,
though this option is much less common. Microsoft operating systems default to using 802.1H for the
AppleTalk protocol suite and IPX, and use RFC 1042 for all other protocols. Most access points now
conform to the Microsoft behavior, and no longer have an option to switch encapsulation type. In
fact, the Microsoft encapsulation selection is so widely supported that it was part of the Wi-Fi
Alliance's certification test suite at one point.
Contention-Based Data Service
The additional features incorporated into 802.11 to add reliability lead to a confusing tangle of rules
about which types of frames are permitted at any point. They also make it more difficult for network
administrators to know which frame exchanges they can expect to see on networks. This section
clarifies the atomic exchanges that move data on an 802.11 LAN. (Most management frames are
announcements to interested parties in the area and transfer information in only one direction.)
The exchanges presented in this section are atomic, which means that they should be viewed as a
single unit. Two distinct sets of atomic exchanges are defined by 802.11. One is used by the DCF for
contention-based service; those exchanges are described in this chapter. A second set of exchanges
is specified for use with the PCF for contention-free services. Frame exchanges used with contentionfree services are intricate and harder to understand. Since very few (if any) commercial products
implement contention-free service, these exchanges are not described.
Frame exchanges under the DCF dominate the 802.11 MAC. According to the rules of the DCF, all
products are required to provide best-effort delivery. To implement the contention-based MAC,
stations process MAC headers for every frame while they are active. Exchanges begin with a station
seizing an idle medium after the DIFS.
Broadcast and Multicast Data or Management Frames
Broadcast and multicast frames, which can also be referred to as group frames because they are
destined for more than one receiving station, have the simplest frame exchanges because there is no
acknowledgment. Framing and addressing are somewhat more complex in 802.11, so the types of
frames that match this rule are the following:
Broadcast data frames with a broadcast address in the Address 1 field
Multicast data frames with a multicast address in the Address 1 field
Broadcast management frames with a broadcast address in the Address 1 field (Beacon, Probe
Request, and IBSS ATIM frames)
Frames destined for group addresses cannot be fragmented and are not acknowledged. The entire
atomic sequence is a single frame, sent according to the rules of the contention-based access
control. After the previous transmission concludes, all stations wait for the DIFS and begin counting
down the random delay intervals in the contention window.
Because the frame exchange is a single-frame sequence, the NAV is set to 0. With no further frames
to follow, there is no need to use the virtual carrier-sense mechanism to lock other stations out of
using the medium. After the frame is transmitted, all stations wait through the DIFS and begin
counting down through the contention window for any deferred frames. See Figure 3-14.
Figure 3-14. Broadcast/multicast data and broadcast management atomic
frame exchange
Depending on the environment, frames sent to group addresses may have lower service quality
because the frames are not acknowledged. Some stations may therefore miss broadcast or multicast
traffic, but there is no facility built into the MAC for retransmitting broadcast or multicast frames.
Unicast Frames
Frames that are destined for a single station are called directed data by the 802.11 standard. This
book uses the more common term unicast. Unicast frames must be acknowledged to ensure
reliability, which means that a variety of mechanisms can be used to improve efficiency. All the
sequences in this section apply to any unicast frame and thus can apply to management frames and
data frames. In practice, these operations are usually observed only with data frames.
Basic positive acknowledgment (final fragment)
Reliable transmission between two stations is based on simple positive acknowledgments. Unicast
data frames must be acknowledged, or the frame is assumed to be lost. The most basic case is a
single frame and its accompanying acknowledgment, as shown in Figure 3-15.
Figure 3-15. Basic positive acknowledgment of data
The frame uses the NAV to reserve the medium for the frame, its acknowledgment, and the
intervening SIFS. By setting a long NAV, the sender locks the virtual carrier for the entire sequence,
guaranteeing that the recipient of the frame can send the acknowledgment. Because the sequence
concludes with the ACK, no further virtual carrier locking is necessary, and the NAV in the ACK is set
to 0.
Fragmentation
Many higher-layer network protocols, including IP, incorporate fragmentation. The disadvantage of
network-layer fragmentation is that reassembly is performed by the final destination; if any of the
fragments are lost, the entire packet must be retransmitted. Link layers may incorporate
fragmentation to boost speed over a single hop with a small MTU.[*] 802.11 can also use
fragmentation to help avoid interference. Radio interference is often in the form of short, high-energy
bursts and is frequently synchronized with the AC power line. Breaking a large frame into small
frames allows a larger percentage of the frames to arrive undamaged. The basic fragmentation
scheme is shown in Figure 3-16.
[*]
This is the approach used by Multi-link PPP (RFC 1990).
The last two frames exchanged are the same as in the previous sequence, and the NAV is set
identically. However, all frames prior to the penultimate frame use the NAV to lock the medium for
the next frame. The first data frame sets the NAV for a long enough period to accommodate its ACK,
the next fragment, and the acknowledgment following the next fragment. To indicate that it is a
fragment, the MAC sets the More Fragments bit in the frame control field to 1. All nonfinal ACKs
continue to extend the lock for the next data fragment and its ACK. Subsequent data frames then
Figure 3-16. Fragmentation
continue to lengthen the NAV to include successive acknowledgments until the final data frame,
which sets the More Fragments bit to 0, and the final ACK, which sets the NAV to 0. No limit is placed
on the number of fragments, but the total frame length must be shorter than any constraint placed
on the exchange by the PHY.
Fragmentation is controlled by the fragmentation threshold parameter in the MAC. Most network card
drivers allow you to configure this parameter. Any frames larger than the fragmentation threshold
are fragmented in an implementation-dependent way. Network administrators can change the
fragmentation threshold to tune network behavior. Higher fragmentation thresholds mean that
frames are delivered with less overhead, but the cost to a lost or damaged frame is much higher
because more data must be discarded and retransmitted. Low fragmentation thresholds have much
higher overhead, but they offer increased robustness in the face of hostile conditions.
RTS/CTS
To guarantee reservation of the medium and uninterrupted data transmission, a station can use the
RTS/CTS exchange. Figure 3-17 shows this process. The RTS/CTS exchange acts exactly like the
initial exchange in the fragmentation case, except that the RTS frame does not carry data. The NAV
in the RTS allows the CTS to complete, and the CTS is used to reserve access for the data frame.
RTS/CTS can be used for all frame exchanges, none of them, or something in between. Like
fragmentation, RTS/CTS behavior is controlled by a threshold set in the driver software. Frames
larger than the threshold are preceded by an RTS/CTS exchange to clear the medium, while smaller
frames are simply transmitted.
Figure 3-17. RTS/CTS lockout
RTS/CTS with fragmentation
In practice, the RTS/CTS exchange is often combined with fragmentation (Figure 3-18). Fragmented
frames are usually quite long and thus benefit from the use of the RTS/CTS procedure to ensure
exclusive access to the medium, free from contention from hidden nodes. Some vendors set the
default fragmentation threshold to be identical to the default RTS/CTS threshold.
Figure 3-18. RTS/CTS with fragmentation
Powersaving Sequences
The most power-hungry components in RF systems are the amplifiers used to boost a signal
immediately prior to transmission and to boost the received signal to an intelligible level immediately
after its reception. 802.11 stations can maximize battery life by shutting down the radio transceiver
and sleeping periodically. During sleeping periods, access points buffer any unicast frames for
sleeping stations. These frames are announced by subsequent Beacon frames. To retrieve buffered
frames, newly awakened stations use PS-Poll frames.
An access point that receives a PS-Poll frame may respond with the requested data immediately, or
at its leisure when circumstances permit. In some cases, the PS-Poll response is determined by the
802.11 chipset vendor in the AP. Some chipset vendors support both modes, while others support
only one. 802.11 only requires that one method be supported; both approaches are equally
standards-compliant.
Immediate response
Access points can respond immediately to the PS-Poll. After a short interframe space, an access point
may transmit the frame. Figure 3-19 shows an implied NAV as a result of the PS-Poll frame. The PSPoll frame contains an Association ID in the Duration/ID field so that the access point can determine
which frames were buffered for the mobile station. However, the MAC specification requires all
stations receiving a PS-Poll to update the NAV with an implied value equal to a short interframe
space and one ACK. Although the NAV is too short for the data frame, the access point acquires that
the medium and all stations defer access for the entire data frame. At the conclusion of the data
frame, the NAV is updated to reflect the value in the header of the data frame.
Figure 3-19. Immediate PS-Poll response
If the buffered frame is large, it may require fragmentation. Figure 3-20 illustrates an immediate PSPoll response requiring fragmentation. Like all other stations, access points typically have a
configurable fragmentation threshold.
Deferred response
Instead of an immediate response, access points can also respond with a simple acknowledgment.
This is called a deferred response because the access point acknowledges the request for the
buffered frame but does not act on it immediately. One of the advantages of using deferred response
is that the software on the access point is somewhat easier to implement because the
acknowledgment can be
Figure 3-20. Immediate PS-Poll response with fragmentation
transmitted immediately by the chipset firmware, and the buffered data may be queued for
transmission normally.
A station requesting a frame with a PS-Poll must stay awake until it is delivered. Under contentionbased service, however, the access point can deliver a frame at any point. A station cannot return to
a low-power mode until it receives a Beacon frame in which its bit in the traffic indication map (TIM)
is clear.
Figure 3-21 illustrates this process. In this figure, the station has recently changed from a low-power
mode to an active mode, and it notes that the access point has buffered frames for it. It transmits a
PS-Poll to the access point to retrieve the buffered frames. However, the access point may choose to
defer its response by transmitting only an ACK. At this point, the access point has acknowledged the
station's request for buffered frames and promised to deliver them at some point in the future. The
station must wait in active mode, perhaps through several atomic frame exchanges, before the
access point delivers the data. A buffered frame may be subject to fragmentation, although Figure 321 does not illustrate this case.
Figure 3-21. Deferred PS-Poll response example
After receiving a data frame, the station must remain awake until the next Beacon is transmitted.
Beacon frames only note whether frames are buffered for a station and have no way of indicating the
number of frames. Once the station receives a Beacon frame indicating that no more traffic is
buffered, it can conclude that it has received the last buffered frame and return to a low-power
mode.
Multirate Support
Network technologies that operate at a variety of different speeds must have a method of negotiating
a mutually acceptable data rate. Speed negotiation is particularly that are available to stations.
Stations may also change speed frequently in response to rapid changes in the radio environment. As
the distance between stations varies, the speed will vary as well. Stations must adapt to the
changing environment by altering transmission speed as necessary. As with many other protocol
features, the 802.11 standards do not specify how a rate is selected. General rules are laid down by
the standard, and vendors have a great deal of freedom in implementing the details. There are a few
general rules that are required of all stations:
1. Every station maintains a list of operational rates, which is the list of rates that are supported
by both the station and the BSS serving it. (A BSS typically corresponds to an access point, but
newer products may offer the ability to customize operational rates on a "virtual AP" basis.) No
frames may be transmitted at a rate that is higher than any rate in the operational rate set.
2. Every BSS must also maintain a basic rate set, which is a list of data rates that must be
supported by every station joining the BSS. Any frame sent to a group receiver address must be
transmitted at a basic rate, ensuring that all stations can demodulate it correctly.
3. Control frames that start a frame exchange, such as RTS and CTS frames, must also be
transmitted at one of the rates in the basic rate set. This rule ensures that a station which must
transmit a CTS frame in response to an RTS frame can do so at the same rate.
Control frames may be used in a backwards-compatiblity mode called protection (see Chapter
14). Protection mode is intended to prevent interference between a station supporting only an
older, slower radio modulation and a newer station that supports faster radio modulations.
Protection frames must be transmitted using older modulations if there are stations incapable of
using newer modulations in the area.
4. Frames destined for a single station will have a unicast destination address in the Address 1
field of the frame. Unicast frames can be transmitted at any rate which is known to be
supported by the destination. Selection of the data rate is not specified by any 802.11 standard.
Frames used in the contention-free period (see Chapter 9) may serve multiple purposes. If a
frame includes an ACK, it is intended for the previous transmitter rather than the frame
recipient. The transmitter must ensure that the frame is transmitted at a rate supported by both
the receiver and the station the ACK is intended for.
5. Response frames, such as acknowledgments or CTS frames, must be transmitted at a rate in
the basic rate set, but no faster than the initial frame in the sequence was transmitted.
Response frames must use the same modulation as the initial frame (DSSS, CCK, or OFDM). [*]
[*]
This is not always strictly obeyed. One chipset vendor always tries to transmit ACK frames at 24 Mbps
because it is usually the highest mandatory data rate.
Rate selection and fallback
Every 802.11 interface on the market supports some sort of rate fallback mechanism, which steps
down the data rate in response to adverse network conditions. Rate selection behavior also
determines when a card determines that it should upgrade its data rate due to improved link quality.
Exactly how an 802.11 station determines it should go to a lower rate (or a higher rate) is not
specified, so the implementation of rate selection is left to chipset vendors. Nearly every chipset has
its own rate selection mechanism, and as a result, most 802.11 interfaces behave differently. Rate
selection behavior is programmable, and is controlled by the code that runs the interface. It may be
altered by driver version changes or firmware changes to an interface.
The most common algorithms used to determine when to change the data rate depend on some
loose measurement of signal quality. Signal quality may be measured directly by the signal-to-noise
ratio, or indirectly by observing how many frames require retransmission. Direct measurements of
signal-to-noise ratios may be an instantaneous measurement of the last frame, or they may be
averaged over some recent period or number of recently received frames. Some chipsets use a direct
measurement of signal-to-noise ratio, but will convert it into a "signal quality" measurement first. As
the signal quality drops, the chip reduces the data rate to compensate.[ ]
[
] The signal quality measurement is often used to make roaming decisions when stations move between APs, though no
standard specifies how or even whether signal quality is an input to the algorithm.
Indirect measurements may monitor instantaneous or average frame loss, and compensate
appropriately. A simple algorithm using indirect measurement is "if the frame is lost and the frame
retry counter is exhausted, step down to the next data rate and try again; repeat until the frame is
delivered or the slowest data rate is unable to successfully deliver." Chipsets that perform indirect
signal quality measurements may make some modifications to avoid a time-consuming step down
through all the rates supported by the PHY, especially since recent chipsets support a great number
of rates, and the time to retry at the slower rates is extremely time-consuming.
Frame Processing and Bridging
At the core, a wireless access point is a glorified bridge that translates frames between a wireless
medium and a wired medium. Although 802.11 does not place any constraint on the wired medium's
technology, I am not aware of an access point that does not use Ethernet. Most access points are
designed as 802.11-to-Ethernet bridges, so it is important to understand the way that frames are
transferred between the two media. See Figure 3-22.
Figure 3-22. Translating frames between a wireless and a wired medium
Wireless Medium to Wired Medium (802.11 to Ethernet)
When a frame is received on the wireless interface of an access point bound for the wired network,
the access point must bridge the frame between the two media. Informally, this is the series of tasks
performed by the access point:
1.
1. When a frame is received at the access point, it first checks for basic integrity. Physical layer
headers that are discussed in the chapters for their respective physical layers are checked, and
the FCS on the 802.11 frame is validated.
2. After verifying that the frame was likely received without error, the access point checks whether
it should process the frame further.
a. Frames sent to an access point have the MAC address of the AP (the BSSID) in the
Address 1 field of the 802.11 MAC header. Frames that do not match the BSSID of the AP
should be discarded. (This step is not implemented by many products.)
b. The 802.11 MAC detects and removes duplicate frames. Frames may be duplicated for a
variety of reasons, but one of the most common is that the 802.11 acknowledgment is lost
or corrupted in transit. To simplify higher-level processing, the 802.11 MAC is responsible
for filtering out duplicate frames.
3. Once the access point determines the frame should be processed further, it must decrypt
frames protected by a link layer security algorithm. Details of the decryption can be found in
subsequent chapters on security.
4. Once the frame is successfully decrypted, it is checked to see if it is a constituent fragment of a
larger frame that requires reassembly. Reassembled frames are subject to integrity protection
on the whole reassembled unit, rather than the individual components.
5. If the AP should bridge the frame, as determined by the BSSID check in step 2a, the relatively
complex wireless MAC header is translated into the simple Ethernet MAC header.
a. The destination address, which is found in the Address 3 field of the 802.11 MAC header,
is copied to the Ethernet destination address.
b. The source address, which is found in the Address 2 field of the 802.11 MAC header, is
copied to the Ethernet source address.
c. The Ethernet type code is copied from the SNAP header in the 802.11 Data field to the
Type code in the Ethernet frame. If the Ethernet frame uses SNAP as well, the entire SNAP
header can be copied.
d. Sequence information is used in fragmentation reassembly, but is discarded when the
frame is bridged.
e. When quality of service processing is standardized, the QoS mapping from the wireless
interface to the wired interface will occur here. For the moment, though, it sufficies to say
that it may take the form of an 802.1p prioritization bit on the wired frame, or some other
form of control.
6. The FCS is recalculated. Both Ethernet and 802.11 use the same algorithm to calculate the FCS,
but the 802.11 frame has several extra fields that are protected by the FCS.
7. The new frame is transmitted on the Ethernet interface.
Wired Medium to Wireless Medium (Ethernet to 802.11)
Bridging frames from the wired side of an access point to the wireless medium is quite similar to the
reverse process:
1. After validating the Ethernet FCS, the access point first checks that a received frame should be
processed further by checking that the destination address belongs to a station currently
associated with the access point.
2. The SNAP header is prepended to the data in the Ethernet frame. The higher-level packet is
encapsulated within a SNAP header whose Type code is copied from the Ethernet type code. If
the Ethernet frame uses SNAP as well, the entire SNAP header can be copied.
3. The frame is scheduled for transmission. 802.11 includes complex powersaving operations that
may cause an access point to hold the frame in a buffer before placing it on the transmit queue.
Powersaving operations are described in Chapter 8.
4. Once a frame has been queued for transmission, it is assigned a sequence number. The
resulting data is protected by an integrity check, if required. If fragmentation is required, the
frame will be split according to the configured fragmentation threshold. If the frame is
fragmented, fragment numbers in the Sequence Control field will also be assigned.
5. If frame protection is required, the body of the frame (or each fragment) is encrypted.
6. The 802.11 MAC header is constructed from the Ethernet MAC header.
a. The Ethernet destination address is copied to the Address 1 field of the 802.11 MAC
header.
b. The BSSID is placed in the Address 2 field of the MAC header, as the sender of the frame
on the wireless medium.
c. The source address of the frame is copied to the Address 3 field of the MAC header.
d. Other fields in the 802.11 MAC header are filled in. The expected transmission time will be
placed in the Duration field and the appropriate flags are filled in in the Frame Control
field.
7. The FCS is recalculated. Both Ethernet and 802.11 use the same algorithm to calculate the FCS,
but the 802.11 frame has several extra fields that are protected by the FCS.
8. The new frame is transmitted on the 802.11 interface.
Quality of Service Extensions
Quality of service extensions may affect the order in which frames are transmitted, but it does not
alter the basic path that a frame takes through the 802.11 MAC. Rather than using a single transmit
queue, the 802.11e quality of service extensions will have multiple transmit queues operating at
steps 4, 5, and 7 of the wired-to-wireless bridging procedure described above. At each of those
steps, frames are processed in a priority order that can be affected by the contents of the frame and
locally configured prioritization rules.
Chapter 4. 802.11 Framing in Detail
Chapter 3 presented the basic frame structure and the fields that comprise it, but it did not go into
detail about the different frame types. Ethernet framing is a simple matter: add a preamble, some
addressing information, and tack on a frame check at the end. 802.11 framing is much more
involved because the wireless medium requires several management features and corresponding
frame types not found in wired networks.
Three major frame types exist. Data frames are the pack horses of 802.11, hauling data from station
to station. Several different data frame flavors can occur, depending on the network. Control frames
are used in conjunction with data frames to perform area-clearing operations, channel acquisition
and carrier-sensing maintenance functions, and positive acknowledgment of received data. Control
and data frames work in conjunction to deliver data reliably from station to station. Management
frames perform supervisory functions; they are used to join and leave wireless networks and move
associations from access point to access point.
This chapter is intended to be a reference. There is only so much life any author can breathe into
framing details, no matter how much effort is expended to make the details interesting. Please feel
free to skip this chapter in its entirety and flip back when you need in-depth information about frame
structure. With rare exception, detailed framing relationships generally do not fall into the category
of "something a network administrator needs to know." This chapter tends to be a bit acronym-heavy
as well, so refer to the glossary at the back of the book if you do not recognize an acronym.
Data Frames
Data frames carry higher-level protocol data in the frame body. Figure 4-1 shows a generic data
frame. Depending on the particular type of data frame, some of the fields in the figure may not be
used.
Figure 4-1. Generic data frame
The different data frame types can be categorized according to function. One such distinction is
between data frames used for contention-based service and those used for contention-free service.
Any frames that appear only in the contention-free period can never be used in an IBSS. Another
possible division is between frames that carry data and frames that perform management functions.
Table 4-1 shows how frames may be divided along these lines. Frames used in contention-free
service are discussed in detail in Chapter 9.
Table 4-1. Categorization of data frame types
Frame type
Contention-based
service
Contention-free
service
Carries
data
Does not carry
data
Data
Data+CF-Ack
Data+CF-Poll
AP only
Data+CF-Ack+CFPoll
AP only
Null
CF-Ack
CF-Poll
AP only
CF-Ack+CF-Poll
AP only
Frame Control
All the bits in the Frame Control field are used according to the rules described in Chapter 3. Frame
Control bits may affect the interpretation of other fields in the MAC header, though. Most notable are
the address fields, which depend on the value of the ToDS and FromDS bits.
Duration
The Duration field carries the value of the Network Allocation Vector (NAV). Access to the medium is
restricted for the time specified by the NAV. Four rules specify the setting for the Duration field in
data frames:
1. Any frames transmitted during the contention-free period set the Duration field to 32,768.
Naturally, this applies to any data frames transmitted during this period.
2. Frames transmitted to a broadcast or multicast destination (Address 1 has the group bit set)
have a duration of 0. Such frames are not part of an atomic exchange and are not
acknowledged by receivers, so contention-based access to the medium can begin after the
conclusion of a broadcast or multicast data frame. The NAV is used to protect access to the
transmission medium for a frame exchange sequence. With no link-layer acknowledgment
following the transmission of a broadcast or multicast frame, there is no need to lock access to
the medium for subsequent frames.
3. If the More Fragments bit in the Frame Control field is 0, no more fragments remain in the
frame. The final fragment need only reserve the medium for its own ACK, at which point
contention-based access resumes. The Duration field is set to the amount of time required for
one short interframe space and the fragment acknowledgment. Figure 4-2 illustrates this
process. The penultimate fragment's Duration field locks access to the medium for the
transmission of the last fragment.
Figure 4-2. Duration setting on final fragment
4. If the More Fragments bit in the Frame Control field is set to 1, more fragments remain. The
Duration field is set to the amount of time required for transmission of two acknowledgments,
plus three short interframe spaces, plus the time required for the next fragment. In essence,
nonfinal fragments set the NAV just like an RTS would (Figure 4-3); for this reason, they are
referred to as a virtual RTS.
Figure 4-3. Duration settings on nonfinal fragment
Addressing and DS Bits
The number and function of the address fields depends on which of the distribution system bits are
set, so the use of the address fields indirectly depends on the type of network deployed. Table 4-2
summarizes the use of the address fields in data frames. The fourth address field is only used by
wireless bridges, and is therefore relatively uncommon.
Table 4-2. Use of the address fields in data frames
Function
ToDS FromDS
Address 1
(receiver)
Address 2
(transmitter)
Address
3
Address
4
IBSS
0
0
DA
SA
BSSID
Not used
To AP (infra.)
1
0
BSSID
SA
DA
Not used
From AP
(infra.)
0
1
DA
BSSID
SA
Not used
WDS (bridge) 1
1
RA
TA
DA
SA
Address 1 indicates the receiver of the frame. In many cases, the receiver is the destination, but not
always. The destination is the station that will process the network-layer packet contained in the
frame; the receiver is the station that will attempt to decode the radio waves into an 802.11 frame.
If Address 1 is set to a broadcast or multicast address, the BSSID is also checked. Stations respond
only to broadcasts and multicasts originating in the same basic service set (BSS); they ignore
broadcasts and multicasts from different BSSIDs.[*] Address 2 is the transmitter address and is used
to send acknowledgments. Transmitters are not necessarily senders. The sender is the frame that
generated the network-layer protocol packet in the frame; the transmitter put the frame on to the
radio link. The Address 3 field is used for filtering by access points and the distribution system, but
the use of the field depends on the particular type of network used.
[*]
Not all cards perform this BSSID filtering correctly. Many products will pass all broadcasts up to higher protocol layers without
validating the BSSID first.
In the case of an IBSS, no access points are used, and no distribution system is present. The
transmitter is the source, and the receiver is the destination. All frames carry the BSSID so that
stations may check broadcasts and multicasts; only stations that belong to the same BSS will process
broadcasts and multicasts. In an IBSS, the BSSID is created by a random-number generator.
802.11 draws a distinction between the source and transmitter and a parallel distinction between the
destination and the receiver. The transmitter sends a frame on to the wireless medium but does not
necessarily create the frame. A similar distinction holds for destination addresses and receiver
addresses. A receiver may be an intermediate destination, but frames are processed by higher
protocol levels only when they reach the destination.
The BSSID
In the PS-Poll frame, the Duration/ID field is an association ID rather than a value used
by the virtual carrier-sensing function. When mobile stations associate with an access
point, the access point assigns a value called the Association ID (AID) from the range 12,007. The AID is used for a variety of purposes that appear throughout this book.
Each BSS is assigned a BSSID, a 48-bit binary identifier that distinguishes it from other
BSSs throughout the network. The major advantage of the BSSID is filtering. Several
distinct 802.11 networks may overlap physically, and there is no reason for one network
to receive link-layer broadcasts from a physically overlapping network.
In an infrastructure BSS, the BSSID is the MAC address of the wireless interface in the
access point creating the BSS. IBSSs must create BSSIDs for networks brought into
existence. To maximize the probability of creating a unique address, 46 random bits are
generated for the BSSID. The Universal/Local bit for the new BSSID is set to 1,
indicating a local address, and the Individual/Group bit is set to 0. For two distinct IBSSs
to create the same BSSID, they would need to generate an identical random 46 bits.
One BSSID is reserved. The all-1s BSSID is the broadcast BSSID. Frames that use the
broadcast BSSID pass through any BSSID filtering in the MAC. BSSID broadcasts are
used only when mobile stations try to locate a network by sending probe requests. In
order for probe frames to detect the existence of a network, they must not be filtered by
the BSSID filter. Probe frames are the only frames allowed to use the broadcast BSSID.
To expand on these distinctions, consider the use of the address fields in infrastructure networks.
Figure 4-4 shows a simple network in which a wireless client is connected to a server through an
802.11 network. Frames sent by the client to the server use the address fields as specified in the
second line of Table 4-2.
Figure 4-4. Address field usage in frames to the distribution system
In the case of frames bound for a destination on the distribution system, the client is both source and
transmitter. The receiver of the wireless frame is the access point, but the access point is only an
intermediate destination. When the frame reaches the access point, it is relayed to the distribution
system to reach the server. Thus, the access point is the receiver, and the (ultimate) destination is
the server. In infrastructure networks, access points create associated BSSs with the address of their
wireless interfaces, which is why the receiver address (Address 1) is set to the BSSID.
When the server replies to the client, frames are transmitted to the client through the access point,
as in Figure 4-5. This scenario corresponds to the third line in Table 4-2.
Figure 4-5. Address field usage in frames from the distribution system
Frames are created by the server, so the server's MAC address is the source address for frames.
When frames are relayed through the access point, the access point uses its wireless interface as the
transmitter address. As in the previous case, the access point's interface address is also the BSSID.
Frames are ultimately sent to the client, which is both the destination and receiver.
The fourth line in Table 4-2 shows the use of the address fields in a wireless distribution system
(WDS), which is sometimes called a wireless bridge. In Figure 4-6, two wired networks are joined by
access points acting as wireless bridges. Frames bound from the client to the server traverse the
802.11 WDS. The source and destination addresses of the wireless frames remain the client and
server addresses. These frames, however, also identify the transmitter and receiver of the frame on
the wireless medium. For frames bound from the client to the server, the transmitter is the clientside access point, and the receiver is the server-side access point. Separating the source from the
transmitter allows the server-side access point to send required 802.11 acknowledgments to its peer
access point without interfering with the wired link layer.
Figure 4-6. Wireless distribution system
Variations on the Data Frame Theme
802.11 uses several different data frame types. Variations depend on whether the service is
contention-based or contention-free. Contention-free frames can incorporate several functions for the
sake of efficiency. Data may be transmitted, but by changing the frame subtype, data frames in the
contention-free period may be used to acknowledge other frames, saving the overhead of interframe
spaces and separate acknowledgments. Here are the different data frame types that are commonly
used:
Data
Frames of the Data subtype are transmitted only during the contention-based access periods.
They are simple frames with the sole purpose of moving the frame body from one station to
another.
Null
Null frames[*] are a bit of an oddity. They consist of a MAC header followed by the FCS trailer.
In a traditional Ethernet, empty frames would be extraneous overhead; in 802. 11 networks,
they are used by mobile stations to inform the access point of changes in power-saving status.
When stations sleep, the access point must begin buffering frames for the sleeping station. If
the mobile station has no data to send through the distribution system, it can use a Null frame
with the Power Management bit in the Frame Control field set. Access points never enter
power-saving mode and do not transmit Null frames. Usage of Null frames is shown in Figure
4-7.
[*]
To indicate that Null is used as the frame type from the specification rather than the English word, it is
capitalized. This convention will be followed throughout the chapter.
Figure 4-7. Data frame of subtype Null
Several other frame types exist for use within the contention-free period. However, contention-free
service is not widely implemented; the discussion of the contention-free frames (Data+CF-Ack,
Data+CF-Poll, Data+CF-Ack+CF-Poll, CF-Ack, CF-Poll, and CF-Ack+CF-Poll) can be found in Chapter
9.
Applied Data Framing
The form of a data frame can depend on the type of network. The actual subtype of the frame is
determined solely by the subtype field, not by the presence or absence of other fields in the frame.
IBSS frames
In an IBSS, three address fields are used, as shown in Figure 4-8. The first address identifies the
receiver, which is also the destination address in an IBSS. The second address is the source address.
After the source and destination addresses, data frames in an IBSS are labeled with the BSSID.
When the wireless MAC receives a frame, it checks the BSSID and passes only frames in the station's
current BSSID to higher protocol layers.
Figure 4-8. IBSS data frame
IBSS data frames have the subtype data or Null; the latter is used only to communicate power
management state.
Frames from the AP
Figure 4-9 shows the format of a frame sent from an access point to a mobile station. As in all data
frames, the first address field indicates the receiver of the frame on the wireless network, which is
the frame's destination. The second address holds the transmitter address. On infrastructure
networks, the transmitter address is the address of the station in the access point, which is also the
BSSID. Finally, the frame indicates the source MAC address of the frame. The split between source
and transmitter is necessary because the 802.11 MAC sends acknowledgments to the frame's
transmitter (the access point), but higher layers send replies to the frame's source.
Figure 4-9. Data frames from the AP
Nothing in the 802.11 specification forbids an access point from transmitting Null frames, but there is
no reason to transmit them. Access points are forbidden from using the power-saving routines, and
they can acknowledge Null frames from stations without using Null frames in response. In practice,
access points send Data frames during the contention-based access period, and they send frames
incorporating the CF-Poll feature during the contention-free period.
Frames to the AP
Figure 4-10 shows the format of a frame sent from a mobile station in an infrastructure network to
the access point currently serving it. The receiver address is the BSSID. In infrastructure networks,
the BSSID is taken from the MAC address of the network station in the access point. Frames destined
for an access point take their source/transmitter address from the network interface in the wireless
station. Access points do not perform filtering, but instead use the third address to forward data to
the appropriate location in the distribution system.
Frames from the distribution system have the ToDS bit set, but the FromDS bit is 0. Mobile stations
in an infrastructure network cannot become the point coordinator, and thus never send frames that
incorporate the contention-free polling (CF-Poll) functions.
Frames in a WDS
When access points are deployed in a wireless bridge (or WDS) topology, all four address fields are
used, as shown in Figure 4-11. Like all other data frames, WDS frames use the first address for the
receiver of the frame and the second address for the transmitter. The MAC uses these two addresses
for acknowledgments and control traffic, such as RTS, CTS, and ACK frames. Two more address fields
are necessary to indicate the source and destination of the frame and distinguish them from the
addresses used on the wireless link.
Figure 4-10. Data frames to the AP
Figure 4-11. WDS frames
On a wireless bridging link, there are no mobile stations, and the contention-free period is not used.
Access points are forbidden to enter power-saving modes, so the power management bit is always
set to 0.
Encrypted frames
Frames protected by link layer security protocols are not new frame types. When a frame is handled
by encryption, the Protected Frame bit in the Frame Control field is set to 1, and the Frame Body field
begins with the appropriate cryptographic header described in Chapters 5 or Chapter 7, depending
on the protocol.
Control Frames
Control frames assist in the delivery of data frames. They administer access to the wireless medium
(but not the medium itself) and provide MAC-layer reliability functions.
Common Frame Control Field
All control frames use the same Frame Control field, which is shown in Figure 4-12.
Figure 4-12. Frame Control field in control frames
Protocol version
The protocol version is shown as 0 in Figure 4-12 because that is currently the only version.
Other versions may exist in the future.
Type
Control frames are assigned the Type identifier 01. By definition, all control frames use this
identifier.
Subtype
This field indicates the subtype of the control frame that is being transmitted.
ToDS and FromDS bits
Control frames arbitrate access to the wireless medium and thus can only originate from
wireless stations. The distribution system does not send or receive control frames, so these bits
are always 0.
More Fragments bit
Control frames are not fragmented, so this bit is always 0.
Retry bit
Control frames are not queued for retransmission like management or data frames, so this bit
is always 0.
Power Management bit
This bit is set to indicate the power management state of the sender after conclusion of the
current frame exchange.
More Data bit
The More Data bit is used only in management and data frames, so this bit is set to 0 in control
frames.
Protected Frame bit
Control frames may not be encrypted. Thus, for control frames, the Protected Frame bit is
always 0.
Order bit
Control frames are used as components of atomic frame exchange operations and thus cannot
be transmitted out of order. Therefore, this bit is set to 0.
Request to Send (RTS)
RTS frames are used to gain control of the medium for the transmission of "large" frames, in which
"large" is defined by the RTS threshold in the network card driver. Access to the medium can be
reserved only for unicast frames; broadcast and multicast frames are simply transmitted. The format
of the RTS frame is shown in Figure 4-13. Like all control frames, the RTS frame is all header. No
data is transmitted in the body, and the FCS immediately follows the header.
Figure 4-13. RTS frame
Four fields make up the MAC header of an RTS:
Frame Control
There is nothing special about the Frame Control field. The frame subtype is set to 1011 to
indicate an RTS frame, but otherwise, it has all the same fields as other control frames. (The
most significant bits in the 802.11 specification come at the end of fields, so bit 7 is the most
significant bit in the subtype field.)
Duration
An RTS frame attempts to reserve the medium for an entire frame exchange, so the sender of
an RTS frame calculates the time needed for the frame exchange sequence after the RTS frame
ends. The entire exchange, which is depicted in Figure 4-14, requires three SIFS periods, the
duration of one CTS, the final ACK, plus the time needed to transmit the frame or first
fragment. (Fragmentation bursts use subsequent fragments to update the Duration field.) The
number of microseconds required for the transmission is calculated and placed in the Duration
field. If the result is fractional, it is rounded up to the next microsecond.
Figure 4-14. Duration field in RTS frame
Address 1: Receiver Address
The address of the station that is the intended recipient of the large frame.
Address 2: Transmitter Address
The address of the sender of the RTS frame.
Clear to Send (CTS)
The CTS frame, whose format is shown in Figure 4-15, has two purposes. Initially, CTS frames were
used only to answer RTS frames, and were never generated without a preceding RTS. CTS frames
were later adopted for use by the 802.11g protection mechanism to avoid interfering with older
stations. The protection mechanism is described with the rest of 802.11g in Chapter 14.
Figure 4-15. CTS frame
Three fields make up the MAC header of a CTS frame:
Frame Control
The frame subtype is set to 1100 to indicate a CTS frame.
Duration
When used in response to an RTS, the sender of a CTS frame uses the duration from the RTS
frame as the basis for its duration calculation. RTS frames reserve the medium for the entire
RTS-CTS-frame-ACK exchange. By the time the CTS frame is transmitted, though, only the
pending frame or fragment and its acknowledgment remain. The sender of a CTS frame
subtracts the time required for the CTS frame and the short interframe space that preceded
the CTS from the duration in the RTS frame, and places the result of that calculation in the
Duration field. Figure 4-16 illustrates the relationship between the CTS duration and the RTS
duration. Rules for CTS frames used in protection exchanges are described with the protection
mechanism.
Address 1: Receiver Address
The receiver of a CTS frame is the transmitter of the previous RTS frame, so the MAC copies
the transmitter address of the RTS frame into the receiver address of the CTS frame. CTS
frames used in 802.11g protection are sent to the sending station, and are used only to set the
NAV.
Figure 4-16. CTS duration
Acknowledgment (ACK)
ACK frames are used to send the positive acknowledgments required by the MAC and are used with
any data transmission, including plain transmissions, frames preceded by an RTS/CTS handshake,
and fragmented frames (see Figure 4-17). Quality-of-service enhancements relax the requirement for
a single acknowledgment per Data frame. To assess the impact of acknowledgments on net
throughput, see Chapter 25.
Figure 4-17. ACK frame
Three fields make up the MAC header of an ACK frame:
Frame Control
The frame subtype is set to 1101 to indicate an ACK frame.
Duration
The duration may be set in one of two ways, depending on the position of the ACK within the
frame exchange. ACKs for complete data frames and final fragments in a fragment burst set
the duration to 0. The data sender indicates the end of a data transmission by setting the More
Fragments bit in the Frame Control header to 0. If the More Fragments bit is 0, the
transmission is complete, and there is no need to extend control over the radio channel for
additional transmissions. Thus, the duration is set to 0.
If the More Fragments bit is 1, a fragment burst is in progress. The Duration field is used like
the Duration field in the CTS frame. The time required to transmit the ACK and its short
interframe space is subtracted from the duration in the most recent fragment (Figure 4-18).
The duration calculation in nonfinal ACK frames is similar to the CTS duration calculation. In
fact, the 802.11 specification refers to the duration setting in the ACK frames as a virtual CTS.
Figure 4-18. Duration in non-final ACK frames
Address 1: Receiver Address
The receiver address is copied from the transmitter of the frame being acknowledged.
Technically, it is copied from the Address 2 field of the frame being acknowledged.
Acknowledgments are transmitted in response to directed data frames, management frames,
and PS-Poll frames.
Power-Save Poll (PS-Poll)
When a mobile station wakes from a power-saving mode, it transmits a PS-Poll frame to the access
point to retrieve any frames buffered while it was in power-saving mode. The format of the PS-Poll
frame is shown in Figure 4-19. Further details on the operation of power saving modes appears in
Chapter 8.
Figure 4-19. PS-Poll frame
Four fields make up the MAC header of a PS-Poll frame:
Frame Control
The frame subtype is set to 1010 to indicate a PS-Poll frame.
Association ID (AID)
Instead of a Duration field, the PS-Poll frame uses the third and fourth bytes in the MAC header
for the association ID. This is a numeric value assigned by the access point to identify the
association. Including this ID in the frame allows the access point to find any frames buffered
for the now-awakened mobile station.
Address 1: BSSID
This field contains the BSSID of the BSS created by the access point that the sender is
currently associated with.
Address 2: Transmitter Address
This is the address of the sender of the PS-Poll frame.
The PS-Poll frame does not include duration information to update the NAV. However, all stations
receiving a PS-Poll frame update the NAV by the short interframe space plus the amount of time
required to transmit an ACK. The automatic NAV update allows the access point to transmit an ACK
with a small probability of collision with a mobile station.
Association ID (AID)
In the PS-Poll frame, the Duration/ID field is an association ID rather than a value used
by the virtual carrier-sensing function. When mobile stations associate with an access
point, the access point assigns a value called the Association ID (AID) from the range 12,007. The AID is used for a variety of purposes that appear throughout this book.
Management Frames
Management is a large component of the 802.11 specification. Several different types of
management frames are used to provide services that are simple on a wired network. Establishing
the identity of a network station is easy on a wired network because network connections require
dragging wires from a central location to the new workstation. In many cases, patch panels in the
wiring closet are used to speed up installation, but the essential point remains: new network
connections can be authenticated by a personal visit when the new connection is brought up.
Wireless networks must create management features to provide similar functionality. 802.11 breaks
the procedure up into three components. Mobile stations in search of connectivity must first locate a
compatible wireless network to use for access. With wired networks, this step typically involves
finding the appropriate data jack on the wall. Next, the network must authenticate mobile stations to
establish that the authenticated identity is allowed to connect to the network. The wired-network
equivalent is provided by the network itself. If signals cannot leave the wire, obtaining physical
access is at least something of an authentication process. Finally, mobile stations must associate with
an access point to gain access to the wired backbone, a step equivalent to plugging the cable into a
wired network.
The Structure of Management Frames
802.11 management frames share the structure shown in Figure 4-20. The MAC header is the same
in all management frames; it does not depend on the frame subtype. Management frames use
information elements, little chunks of data with a numerical label, to communicate information to
other systems.
Figure 4-20. Generic management frame
Address fields
As with all other frames, the first address field is used for the frame's destination address. Some
management frames are used to maintain properties within a single BSS. To limit the effect of
broadcast and multicast management frames, stations are required to inspect the BSSID after
receiving a mangement frame, though not all implementations perform BSSID filtering. Only
broadcast and multicast frames from the BSSID that a station is currently associated with are passed
to MAC management layers. The one exception to this rule is Beacon frames, which are used to
announce the existence of an 802.11 network.
BSSIDs are assigned in the familiar manner. Access points use the MAC address of the wireless
network interface as the BSSID. Mobile stations adopt the BSSID of the access point they are
currently associated with. Stations in an IBSS use the randomly generated BSSID from the BSS
creation. One exception to the rule: frames sent by the mobile station seeking a specific network
may use the BSSID of the network they are seeking, or they may use the broadcast BSSID to find all
networks in the vicinity.
Duration calculations
Management frames use the Duration field in the same manner that other frames do:
1. Any frames transmitted in the contention-free period set the duration to 32,768.
2. Frames transmitted during the contention-based access periods using only the DCF use the
Duration field to block access to the medium to allow any atomic frame exchanges to complete.
a. If the frame is a broadcast or multicast (the destination address is a group address), the
duration is 0. Broadcast and multicast frames are not acknowledged, so the NAV is not
needed to block access to the medium.
b. If a nonfinal fragment is part of a multiframe exchange, the duration is set to the number
of microseconds taken up by three SIFS intervals plus the next fragment and its
acknowledgment.
c. Final fragments use a duration that is the time required for one acknowledgment plus one
SIFS.
Frame body
Management frames are quite flexible. Most of the data contained in the frame body uses fixedlength fields called fixed fields and variable-length fields called information elements. Information
elements are blobs of data of varying size. Each data blob is tagged with a type number and a size,
and it is understood that an information element of a certain type has its data field interpreted in a
certain way. New information elements can be defined by newer revisions to the 802.11
specification; implementations that predate the revisions can ignore newer elements. Old
implementations depend on backward-compatible hardware and frequently can't join networks based
on the newer standards. Fortunately, new options usually can be easily turned off for compatibility.
This section presents the fixed fields and information elements as building blocks and shows how the
building blocks are assembled into management frames. 802.11 mandates the order in which
information elements appear, but not all elements are mandatory. This book shows all the frame
building blocks in the specified order, and the discussion of each subtype notes which elements are
rare and which are mutually exclusive.
Fixed-Length Management Frame Components
10 fixed-length fields may appear in management frames. Fixed-length fields are often referred to
simply as fields to distinguish them from the variable-length information elements. Fields do not have
a header to distinguish them from other parts of the frame body. Because they have a fixed length
and appear in a known order, fields can be delimited without using a field header.
Authentication Algorithm Number
Two bytes are used for the Authentication Algorithm Number field, which are shown in Figure 4-21.
This field identifies the type of authentication used in the initial 802.11-level authentication process
before association occurs. 802.1X authentication occurs after association, and is not assigned an
algorithm number. (The authentication process is discussed more thoroughly in Chapter 8.) The
values permitted for this field are shown in Table 4-3. Only two values are currently defined. Other
values are reserved for future standardization work.
Figure 4-21. Authentication Algorithm Number field
Table 4-3. Values of the Authentication Algorithm Number field
Value
Meaning
0
Open System authentication (typically used with 802.1X authentication)
1
Shared Key authentication (deprecated by 802.11i)
2-65,535 Reserved
Authentication Transaction Sequence Number
Authentication is a multistep process that consists of a challenge from the access point and a
response from the mobile station attempting to associate. The Authentication Transaction Sequence
Number, shown in Figure 4-22, is a two-byte field used to track progress through the authentication
exchange. It takes values from 1 to 65,535; it is never set to 0. Use of this field is discussed in
Chapter 8.
Figure 4-22. Authentication Transaction Sequence Number field
Beacon interval
Beacon transmissions announce the existence of an 802.11 network at regular intervals. Beacon
frames carry information about the BSS parameters and the frames buffered by access points, so
mobile stations must listen to Beacons. The Beacon Interval, shown in Figure 4-23, is a 16-bit field
set to the number of time units between Beacon transmissions. One time unit, which is often
abbreviated TU, is 1,024 microseconds (ms), which is about 1 millisecond. [*] Time units may also be
called kilo-microseconds in various documentation (Kms or kms). It is common for the Beacon
interval to be set to 100 time units, which corresponds to an interval between Beacon transmissions
of approximately 100 milliseconds or 0.1 seconds.
[*]
Kilo-microseconds are an odd blend of the powers-of-2 used in computing for the kilo, and the more common 1/1,000 for micro.
Presumably, the International Bureau of Weights and Measures would protest the mangling of the traditional form of the prefixes.
Figure 4-23. Beacon Interval field
Capability Information
The 16-bit Capability Information field (Figure 4-24) is used in Beacon transmissions to advertise the
network's capabilities. Capability Information is also used in Probe Request and Probe Response
frames. In this field, each bit is used as a flag to advertise a particular function of the network.
Stations use the capability advertisement to determine whether they can support all the features in
the BSS. Stations that do not implement all the features in the capability advertisement are not
allowed to join.
Figure 4-24. Capability Information field
ESS/IBSS
These two bits are mutually exclusive. Access points set the ESS field to 1 and the IBSS field to
0 to indicate that the access point is part of an infrastructure network. Stations in an IBSS set
the ESS field to 0 and the IBSS field to 1.
Privacy
Setting the Privacy bit to 1 requires the use of WEP for confidentiality. In infrastructure
networks, the transmitter is an access point. In IBSSs, Beacon transmission must be handled
by a station in the IBSS.
Short Preamble
This field was added to 802.11b to support the high-rate DSSS PHY. Setting it to 1 indicates
that the network is using the short preamble as described in Chapter 12. Zero means the
option is not in use and is forbidden in the BSS. 802.11g requires use of the short preamble, so
this field is always set to 1 in a network built on the 802.11g standard.
PBCC
This field was added to 802.11b to support the high-rate DSSS PHY. When it is set to 1, it
indicates that the network is using the packet binary convolution coding modulation scheme
described in Chapter 12, or the higher-speed 802.11g PBCC modulation described in Chapter
14. Zero means that the option is not in use and is forbidden in the BSS.
Channel Agility
This field was added to 802.11b to support the high rate DSSS PHY. When it is set to one, it
indicates that the network is using the Channel Agility option described in Chapter 12. Zero
means the option is not in use and is forbidden in the BSS.
Short Slot Time (802.11g)
This bit is set to one to indicate the use of the shorter slot time supported by 802.11g, which is
discussed in Chapter 14.
DSSS-OFDM (802.11g)
This bit is set to one to indicate that the optional DSSS-OFDM frame construction in 802.11g is
in use.
Contention-free polling bits
Stations and access points use these two bits as a label. The meanings of the labels are shown
in Table 4-4.
Table 4-4. Interpretation of polling bits in Capability Information
CF-Poll
Request
Interpretation
0
0
Station does not support polling
0
1
Station supports polling but does not request to be put on the polling
list
1
0
Station supports polling and requests a position on the polling list
1
1
Station supports polling and requests that it never be polled (results in
station treated as if it does not support contention-free operation)
0
0
Access point does not implement the point coordination function
0
1
Access point uses PCF for delivery but does not support polling
1
0
Access point uses PCF for delivery and polling
1
1
Reserved; unused
CF-Pollable
Station
usage
Access point
usage
Current AP Address
Mobile stations use the Current AP Address field, shown in Figure 4-25, to indicate the MAC address
of the access point with which they are associated. This field is used to ease associations and
reassociations. Stations transmit the address of the access point that handled the last association
with the network. When an association is established with a different access point, this field can be
used to transfer the association and retrieve any buffered frames.
Figure 4-25. Current AP Address field
Listen interval
To save battery power, stations may shut off the antenna units in 802.11 network interfaces. While
stations are sleeping, access points must buffer frames for them. Dozing stations periodically wake
up to listen to traffic announcements to determine whether the access point has any buffered frames.
When stations associate with an access point, part of the saved data is the Listen Interval, which is
the number of Beacon intervals that stations wait between listening for Beacon frames. The Listen
Interval, shown in Figure 4-26, allows mobile stations to indicate how long the access point must
retain buffered frames. Higher listen intervals require more access point memory for frame buffering.
Access points may use this feature to estimate the resources that will be required and may refuse
resource-intensive associations. The Listen Interval is described in Chapter 8.
Figure 4-26. Listen Interval field
Association ID
The Association ID, shown in Figure 4-27, is a 16-bit field. When stations associate with an access
point, they are assigned an Association ID to assist with control and management functions. Even
though 14 bits are available for use in creating Association IDs, they range only from 1-2,007. To
maintain compatibility with the Duration/ID field in the MAC header, the two most significant bits are
set to 1.
Timestamp
The Timestamp field, shown in Figure 4-28, allows synchronization between the stations in a BSS.
The master timekeeper for a BSS periodically transmits the number of microseconds it has been
active. When the counter reaches its maximum value, it wraps around. (Counter wraps are unlikely
given the length of time it takes to wrap a
Figure 4-27. Association ID field
64-bit counter. At over 580,000 years, I would bet on a required patch or two before the counter
wrap.)
Figure 4-28. Timestamp field
Reason Code
Stations may send Disassociation or Deauthentication frames in response to traffic when the sender
has not properly joined the network. Part of the frame is a 16-bit Reason Code field, shown in Figure
4-29, to indicate what the sender has done incorrectly. Table 4-5 shows why certain reason codes
are generated. Fully understanding the use of reason codes requires an understanding of the
different classes of frames and states of the 802.11 station, which is discussed in the section "Frame
Transmission and Association and Authentication States."
Figure 4-29. Reason Code field
Table 4-5. Reason codes
Code
Explanation
0
Reserved; unused
1
Unspecified
2
Prior authentication is not valid
3
Station has left the basic service area or extended service area and is deauthenticated
4
Inactivity timer expired and station was disassociated
5
Disassociated due to insufficient resources at the access point
6
Incorrect frame type or subtype received from unauthenticated station
7
Incorrect frame type or subtype received from unassociated station
8
Station has left the basic service area or extended service area and is disassociated
9
Association or reassociation requested before authentication is complete
Code
Explanation
10
(802.11h)
Disassociated because of unacceptable values in Power Capability element
11
(802.11h)
Disassociated because of unacceptable values in Supported Channels element
12
Reserved
13 (802.11i) Invalid information element (added with 802.11i, and likely one of the 802.11i
information elements)
14 (802.11i) Message integrity check failure
15 (802.11i) 4-way keying handshake timeout
16 (802.11i) Group key handshake timeout
17 (802.11i) 4-way handshake information element has different security parameters from initial
parameter set
18 (802.11i) Invalid group cipher
19 (802.11i) Invalid pairwise cipher
20 (802.11i) Invalid Authentication and Key Management Protocol
21 (802.11i) Unsupported Robust Security Network Information Element (RSN IE) version
22 (802.11i) Invalid capabilities in RSN information element
23 (802.11i) 802.1X authentication failure
24 (802.11i) Proposed cipher suite rejected due to configured policy
25-65,535
Reserved; unused
Status Code
Status codes indicate the success or failure of an operation. The Status Code field, shown in Figure 430, is 0 when an operation succeeds and nonzero on failure. Table 4-6 shows the status codes that
have been standardized.
Figure 4-30. Status Code field
Table 4-6. Status codes
Code
Explanation
0
Operation completed successfully
1
Unspecified failure
2-9
Reserved; unused
10
Requested capability set is too broad and cannot be supported
11
Reassociation denied; prior association cannot be identified and transferred
12
Association denied for a reason not specified in the 802.11 standard
13
Requested authentication algorithm not supported
14
Unexpected authentication sequence number
15
Authentication rejected; the response to the challenge failed
16
Authentication rejected; the next frame in the sequence did not arrive in the expected
window
17
Association denied; the access point is resource-constrained
18
Association denied; the mobile station does not support all of the data rates required
by the BSS
19
(802.11b)
Association denied; the mobile station does not support the Short Preamble option
20
(802.11b)
Association denied; the mobile station does not support the PBCC modulation option
21
(802.11b)
Association denied; the mobile station does not support the Channel Agility option
22
(802.11h)
Association denied; Spectrum Management is required
23
(802.11h)
Association denied; Power Capability value is not acceptable
24
(802.11h)
Association denied; Supported Channels is not acceptable
25
(802.11g)
Association denied; the mobile station does not support the Short Slot Time
26
(802.11g)
Association denied; the mobile station does not support DSSS-OFDM
27-39
Reserved
40 (802.11i) Information element not valid
41 (802.11i) Group (broadcast/multicast) cipher not valid
42 (802.11i) Pairwise (unicast) cipher not valid
43 (802.11i) Authentication and Key Management Protocol (AKMP) not valid
44 (802.11i) Robust Security Network information element (RSN IE) version is not supported
45 (802.11i) RSN IE capabilites are not supported
46 (802.11i) Cipher suite rejected due to policy
47-65,535
Reserved for future standardization work
Management Frame Information Elements
Information elements are variable-length components of management frames. A generic information
element has an ID number, a length, and a variable-length component, as shown in Figure 4-31.
Standardized values for the element ID number are shown in Table 4-7.
Figure 4-31. Generic management frame information element
Table 4-7. Information elements
Element ID Name
0
Service Set Identity (SSID)
1
Supported Rates
2
FH Parameter Set
3
DS Parameter Set
4
CF Parameter Set
5
Traffic Indication Map (TIM)
6
IBSS Parameter Set
7 (802.11d) Country
8 (802.11d) Hopping Pattern Parameters
9 (802.11d) Hopping Pattern Table
10
(802.11d)
Request
11-15
Reserved; unused
16
Challenge text
17-31
Reserveda (formerly for challenge text extension, before 802.11 shared key
authentication was discontinued)
32
(802.11h)
Power Constraint
33
(802.11h)
Power Capability
Element ID Name
34
(802.11h)
Transmit Power Control (TPC) Request
35
(802.11h)
TPC Report
36
(802.11h)
Supported Channels
37
(802.11h)
Channel Switch Announcement
38
(802.11h)
Measurement Request
39
(802.11h)
Measurement Report
40
(802.11h)
Quiet
41
(802.11h)
IBSS DFS
42
(802.11g)
ERP information
43-49
Reserved
48 (802.11i) Robust Security Network
50
(802.11g)
Extended Supported Rates
32-255
Reserved; unused
221b
Wi-Fi Protected Access
a
802.11 shared key authentication is no longer recommended, so it is unlikely that these fields will
ever be used.
a
This is used by WPA, and is not an official part of 802.11. However, it is widely implemented, so I
include it in the table.
Service Set Identity (SSID)
Network managers are only human, and they usually prefer to work with letters, numbers, and
names rather than 48-bit identifiers. 802.11 networks, in the broadest sense, are either extended
service sets or independent BSSs. The SSID, shown in Figure 4-32, allows network managers to
assign an identifier to the service set. Stations attempting to join a network may scan an area for
available networks and join the network with a specified SSID. The SSID is the same for all the basic
service areas composing an extended service area.
Figure 4-32. Service Set Identity information element
Some documentation refers to the SSID as the network name because network administrators
frequently assign a character string to it. However, the SSID is just a string of bytes that labels the
BSSID as belonging to a larger agglomeration. Some products require that the string be a garden
variety ASCII string, though the standard has no requirement on the content of the string.
In all cases, the length of the SSID ranges between 0 and 32 bytes. The zero-byte case is a special
case called the broadcast SSID; it is used only in Probe Request frames when a station attempts to
discover all the 802.11 networks in its area.
Supported Rates
Several data rates have been standardized for wireless LANs. The Supported Rates information
element allows an 802.11 network to specify the data rates it supports. When mobile stations
attempt to join the network, they check the data rates used in the network. Some rates are
mandatory and must be supported by the mobile station, while others are optional.
The Supported Rates information element is shown in Figure 4-33. It consists of a string of bytes.
Each byte uses the seven low-order bits for the data rate; the most significant bit indicates whether
the data rate is mandatory. Mandatory rates are encoded with the most significant bit set to 1 and
optional rates have a 0. Up to eight rates may be encoded in the information element. As the number
of data rates has proliferated, the Extended Supported Rates element was standardized to handle
more than eight data rates.
In the initial revision of the 802.11 specification, the seven bits encoded the data rate as a multiple
of 500 kbps. New technology, especially ETSI's HIPERLAN efforts, required a change to the
interpretation. When 7 bits are used to have a multiple of
Figure 4-33. Supported Rates information element
500 kbps, the maximum data rate that can be encoded is 63.5 Mbps. Research and development on
wireless LAN technology has made this rate achievable in the near future. As a result, the IEEE
changed the interpretation from a multiple of 500 kbps to a simple label in 802.11b. Previously
standardized rates were given labels corresponding to the multiple of 500 kbps, but future standards
may use any value. Current standardized values are shown in Table 4-8.
Table 4-8. Supported Rate labels
Binary value
Corresponding rate (Mbps)
2
1
4
2
11 (802.11b)
5.5
12 (802.11g)
6
18 (802.11g)
9
22 (802.11b)
11
24 (802.11g)
12
36 (802.11g)
18
44 (802.11g)
22 (optional 802.11g PBCC)
48 (802.11g)
24
66 (802.11g)
33 (optional 802.11g PBCC)
72 (802.11g)
36
96 (802.11g)
48
108 (802.11g)
54
As an example, Figure 4-33 shows the encoding of two data rates. 2-Mbps service is mandatory and
11-Mbps service is supported. This is encoded as a mandatory 2-Mbps rate and an optional 11-Mbps
rate.
FH Parameter Set
The FH Parameter Set information element, shown in Figure 4-34, contains all parameters necessary
to join a frequency-hopping 802.11 network.
Figure 4-34. FH Parameter Set information element
The FH Parameter Set has four fields that uniquely specify an 802.11 network based on frequency
hopping. Chapter 12 describes these identifiers in depth.
Dwell Time
802.11 FH networks hop from channel to channel. The amount of time spent on each channel
in the hopping sequence is called the dwell time. It is expressed in time units (TUs).
Hop Set
Several hopping patterns are defined by the 802.11 frequency-hopping PHY. This field, a single
byte, identifies the set of hop patterns in use.
Hop Pattern
Stations select one of the hopping patterns from the set. This field, also a single byte, identifies
the hopping pattern in use.
Hop Index
Each pattern consists of a long sequence of channel hops. This field, a single byte, identifies
the current point in the hop sequence.
DS Parameter Set
Direct-sequence 802.11 networks have only one parameter: the channel number used by the
network. High-rate direct sequence networks use the same channels and thus can use the same
parameter set. The channel number is encoded as a single byte, as shown in Figure 4-35.
Figure 4-35. DS Parameter Set information element
Traffic Indication Map (TIM)
Access points buffer frames for mobile stations sleeping in low-power mode. Periodically, the access
point attempts to deliver buffered frames to sleeping stations. A practical reason for this
arrangement is that much more power is required to power up a transmitter than to simply turn on a
receiver. The designers of 802.11 envisioned battery-powered mobile stations; the decision to have
buffered frames delivered to stations periodically was a way to extend battery life for low-power
devices.
Part of this operation is to send the Traffic Indication Map (TIM) information element (Figure 4-36) to
the network to indicate which stations have buffered traffic waiting to be picked up.
Figure 4-36. Traffic Indication Map information element
The meat of the traffic indication map is the virtual bitmap, a logical structure composed of 2,008
bits. Each bit is tied to the Association ID. When traffic is buffered for that Association ID, the bit is
1. If no traffic is buffered, the bit tied to the Association ID is 0.
DTIM Count
This one-byte field is the number of Beacons that will be transmitted before the next DTIM
frame. DTIM frames indicate that buffered broadcast and multicast frames will be delivered
shortly. Not all Beacon frames are DTIM frames.
DTIM Period
This one-byte field indicates the number of Beacon intervals between DTIM frames. Zero is
reserved and is not used. The DTIM count cycles through from the period down to 0.
Bitmap Control and Partial Virtual Bitmap
The Bitmap Control field is divided into two subfields. Bit 0 is used for the traffic indication
status of Association ID 0, which is reserved for multicast traffic. The remaining seven bits of
the Bitmap Control field are used for the Bitmap Offset field.
To save transmission capacity, the Bitmap Offset field can be used to transmit a portion of the
virtual bitmap. The Bitmap Offset is related to the start of the virtual bitmap. By using the
Bitmap Offset and the Length, 802.11 stations can infer which part of the virtual bitmap is
included.
CF Parameter Set
The CF Parameter Set information element is transmitted in Beacons by access points that support
contention-free operation. Contention-free service is discussed in Chapter 9 because of its optional
nature.
IBSS Parameter Set
IBSSs currently have only one parameter, the announcement traffic indication map (ATIM) window,
shown in Figure 4-37. This field is used only in IBSS Beacon frames. It indicates the number of time
units (TUs) between ATIM frames in an IBSS.
Figure 4-37. IBSS Parameter Set information element
Country
The initial 802.11 specifications were designed around the existing regulatory constraints in place in
the major industrialized countries. Rather than continue to revise the specification each time a new
country was added, a new specification was added that provides a way for networks to describe
regulatory constraints to new stations. The main pillar of this is the Country information element,
shown in Figure 4-38.
Figure 4-38. Country information element
After the initial type/length information element header, there is a country identifier, followed by a
series of three-byte descriptors for regulatory constraints. Each constraint descriptor specifies a
unique band, and they may not overlap, since a given frequency has only one maximum allowed
power.
Country String (3 bytes)
A three-character ASCII string of where the station is operating. The first two letters are the
ISO country code (e.g., "US" for the United States). Many countries have different indoor and
outdoor regulations, and the third character distinguishes between the two. When a single set
of omnibus regulations covers all environments, the third character is a space. To designate
indoor or outdoor regulations only, the third character may be set to "I" or "O", respectively.
First Channel Number (1 byte)
The first channel number is the lowest channel subject to the power constraint. Channel
number assignment for each PHY is discussed in the appropriate chapter.
Number of Channels (1 byte)
The size of the band subject to the power constraint is indicated by the number of channels.
The size of a channel is PHY-dependent.
Maximum Transmit Power (1 byte)
The maximum transmit power, expressed in dBm.
Pad (1 byte; optional)
The size of the information element must be an even number of bytes. If the length of the
information element is an odd number of bytes, a single byte of zeroes is appended as a pad.
Hopping Pattern Parameters and Hopping Pattern Table
The initial 802.11 frequency hopping specification, described in Chapter 11, was built around the
regulatory constraints in effect during its design. These two elements can be used to build a hopping
pattern that complies with regulatory constraints in additional countries, which allows further
adoption of the frequency-hopping PHY without requiring additional revision to the specification.
Request
In Probe Request frames, the Request information element is used to ask the network for certain
information elements. The Request information element has the type/length header, and is followed
by a list of integers with the numbers of the information elements being requested (Figure 4-39).
Figure 4-39. Request information element
Challenge Text
The shared-key authentication system defined by 802.11 requires that the mobile station successfully
decrypt an encrypted challenge. The challenge is sent using the Challenge Text information element,
which is shown in Figure 4-40.
Power Constraint
The Power Constraint information element is used to allow a network to describe the maximum
transmit power to stations. In addition to a regulatory maximum, there may be another maximum in
effect. The only field, a one-byte integer, is the number
Figure 4-40. Challenge Text information element
of decibels by which any local constraint reduces the regulatory maximum. If, for example, the
regulatory maximum power were 10 dBm, but this information element contained the value 2, then
the station would set its maximum transmit power to 8 dBm (Figure 4-41).
Figure 4-41. Power Constraint information element
Power Capability
802.11 stations are battery powered, and often have radios that are not as capable as access points,
in part because there is not usually the need for mobile client devices to transmit at high power. The
Power Capability information element allows a station to report its minimum and maximum transmit
power, in integer units of dBm (Figure 4-42).
Figure 4-42. Power Capability information element
TPC Request
The Transmit Power Control (TPC) Request information element is used to request radio link
management information. It has no associated data, so the length field is always zero (Figure 4-43).
Figure 4-43. Transmit Power Request information element
TPC Report
For stations to know how to tune transmission power, it helps to know the attenuation on the link.
TPC Report information elements are included in several types of management frames, and include
two one-byte fields (Figure 4-44). The first, the transmit power, is the transmit power of the frame
containing the information element, in units of dBm. The second, the link margin, represents the
number of decibels of safety that the station requires. Both are used by the station to adapt its
transmission power, as described in Chapter 8.
Figure 4-44. Transmit Power Report information element
Supported Channels
The Supported Channels information element is similar to the Country information element, in that it
describes sub-bands that are supported. After the header, there is a series of sub-band descriptors.
Each sub-band descriptor consists of a first channel number, which is the lowest channel in a
supported sub-band, followed by the number of channels in the sub-band (Figure 4-45). For
example, a device that only supported channels 40 through 52 would set the first channel number to
40, and the number of channels to 12.
Figure 4-45. Supported Channels information element
Channel Switch Announcement
802.11h added the ability of networks to dynamically switch channels. To warn stations in the
network about the impending channel change, management frames may include the Channel Switch
Announcement element shown in Figure 4-46.
Figure 4-46. Channel Switch Announcement information element
Channel Switch Mode
When the operating channel is changed, it disrupts communication. If this field is set to 1,
associated stations should stop transmitting frames until the channel switch has occurred. If it
is set to zero, there is no restriction on frame transmission.
New Channel Number
The new channel number after the switch. At present, there is no need for this field to exceed a
value of 255.
Channel Switch Count
Channel switching can be scheduled. This field is the number of Beacon frame transmission
intervals that it will take to change the channel. Channel switch occurs just before the Beacon
transmission is to begin. A non-zero value indicates the number of Beacon intervals to wait; a
zero indicates that the channel switch may occur without any further warning.
Measurement Request and Measurement Report
Regular channel measurements are important to monitoring the channel and power settings. Two
information elements are defined to allow stations to request measurements and receive reports.
Reports are a key component of 802.11h, and will be discussed in detail in the "Spectrum
Management" section of Chapter 8.
Quiet
One of the reasons for the development of dynamic frequency selection was the need to avoid certain
military radar technologies. To find the presence of radar or other interference, an AP can use the
Quiet element, shown in Figure 4-47, to temporarily shut down the channel to improve the quality of
measurements.
Figure 4-47. Quiet information element
Following the header, there are four fields:
Quiet Count
Quiet periods are scheduled. The count is the number of Beacon transmission intervals until the
quiet period begins. It works in a similar fashion to the Channel Switch Count field.
Quiet Period
Quiet periods may also be periodically scheduled. If this field is zero, it indicates there are no
scheduled quiet periods. A non-zero value indicates the number of beacon intervals between
quiet periods.
Quiet Duration
Quiet periods do not need to last for an entire Beacon interval. This field specifies the number
of time units the quiet period lasts.
Quiet Offset
Quiet periods do not necessarily have to begin with a Beacon interval. The Offset field is the
number of time units after a Beacon interval that the next quiet period will begin. Naturally, it
must be less than one Beacon interval.
IBSS DFS
In an infrastructure network, the access point is responsible for dynamic frequency selection.
Independent networks must have a designated owner of the dynamic frequency selection (DFS)
algorithm. Management frames from the designated station in an IBSS may transmit the IBSS DFS
information element, shown in Figure 4-48.
Figure 4-48. IBSS Dynamic Frequency Selection (DFS) information
element
After the header, it has the MAC address of the station responsible for maintaining DFS information,
as well as a measurement interval. The bulk of the frame is a series of channel maps, which report
what is detected on each channel. The channel map consists of a channel number, followed by a map
byte, which has the following fields:
BSS (1 bit)
This bit will be set if frames from another network are detected during a measurement period.
OFDM Preamble (1 bit)
This bit is set if the 802.11a short training sequence is detected, but without being followed by
the rest of the frame. HIPERLAN/2 networks use the same preamble, but obviously not the
same frame construction.
Unidentified Signal (1 bit)
This bit is set when the received power is high, but the signal cannot be classified as either
another 802.11 network (and hence, set the BSS bit), another OFDM network (and hence, set
the OFDM Preamble bit), or a radar signal (and hence, set the Radar bit). The standard does
not specify what power level is high enough to trigger this bit being set.
Radar (1 bit)
If a radar signal is detected during a measurement period, this bit will be set. Radar systems
which must be detected are defined by regulators, not the 802.11 task group.
Unmeasured (1 bit)
If the channel was not measured, this bit will be set. Naturally, when there was no
measurement taken, nothing can be detected in the band and the previous four bits will be set
to zero.
ERP Information
802.11g defined the extended rate PHY (ERP). To provide backwards compatibility, the ERP
information element, shown in Figure 4-49, was defined. In its first iteration, it is three bit flags in a
single byte.
Non-ERP present
This bit will be set when an older, non-802.11g station associates to a network. It may also be
set when overlapping networks that are not capable of using 802.11g are detected.
Use Protection
When stations incapable of operating at 802.11g data rates are present, the protection bit is
set to 1. This enables backwards compatibility with older stations, as described in Chapter 14.
Barker Preamble Mode
This bit will be set if the stations which have associated to the network are not capable of the
short preamble mode described in Chapter 12.
Figure 4-49. ERP information element
Robust Security Network
With the significant security enhancements in 802.11i, it was necessary to develop a way to
communicate security information between stations. The main tool for this is the Robust Security
Network (RSN) information element, shown in Figure 4-50. There are several variable components,
and in some cases, the RSN information element might run into the limits of the information element
size of 255 bytes past the header.
Version
The version field must be present. 802.11i defined version 1. Zero is reserved, and versions of
two or greater are not yet defined.
Figure 4-50. Robust Security Network (RSN) information element
Group cipher suite
Following the version number is the group cipher suite descriptor. Access points must select a
single group cipher compatible with all associated stations to protect broadcast and multicast
frames. Only one group cipher is allowed.
A cipher suite selector is four bytes long. It starts with an OUI for the vendor, and a number to
identify the cipher. Table 4-9 shows the standardized cipher suites. (Values not shown in the
table are reserved.) The OUI used by 802.11i is 00-0F-AC, which is used by the 802.11
working group.
Table 4-9. Cipher suites
OUI
Suite Type Definition
00-0F-AC (802.11) 0
Use the group cipher suite (only valid for pairwise ciphers)
00-0F-AC
1
WEP-40
00-0F-AC
2
TKIP
OUI
Suite Type Definition
00-0F-AC
3
Reserved
00-0F-AC
4
CCMPa
00-0F-AC
5
WEP-104
Vendor OUI
Any value
Defined by vendor
a
This is the default value for an 802.11i network.
Pairwise Cipher Suites (count + list)
Following the group cipher suite may be several pairwise cipher suites to protect unicast
frames. There is a two-byte count, followed by a series of supported cipher descriptors. The
suite selector may be set to zero to indicate support for only the group cipher suite. There is no
limit, other than the size of the information element, on the number of supported pairwise
ciphers.
Authentication and Key Management (AKM) suites (count + list)
Like the pairwise cipher suite selector, there may be multiple authentication types defined.
Following a count, there is a series of four-byte suite identifiers. As with the cipher suites, the
four-byte identifier consists of an OUI and a suite type number. Table 4-10 shows the standard
authentication types.
Table 4-10. Authentication and key management suites
Suite
type
OUI
Authentication
Key management
00-0F-AC 1
802.1X or PMK
caching
Key derivation from preshared master key, as
described in Chapter 7
00-0F-AC 2
Pre-shared key
Key derivation from pre-shared key, as described
in Chapter 7
Vendor
OUI
Vendor-specific
Vendor-specific
Any
RSN Capabilties
This two-byte field consists of four flags used to describe what the transmitter is capable of,
followed by reserved bits that must be set to zero.
Pre-authentication
An AP may set this bit to indicate it can perform pre-authentication with other APs on the
network to move security sessions around. Otherwise, this bit is set to zero.
Preauthentication is discussed in Chapter 8.
No Pairwise
This bit is set when a station can support a manual WEP key for broadcast data in
conjunction with a stronger unicast key. Although supported by the standard, this
configuration should not be used unless absolutely necessary.
Pairwise Replay Counter and Group Replay Counter
Separate replay counters may be maintained for each priority level defined in emerging
quality of service extensions. These bits describe the number of replay counters
supported by the station.
PMK list (count + list)
Faster hand-offs between access points are possible when the pairwise master key is cached by
the AP. Stations may provide a list of master keys to an AP on association in an attempt to
bypass the time-consuming authentication. PMK caching is discussed in more detail in Chapter
8.
Extended Supported Rates
The Extended Supported Rates information element acts identically to the Supported Rates element
in Figure 4-33, but it allows an information element body of up to 255 bytes to be supported.
Wi-Fi Protected Access (WPA)
Wi-Fi Protected Access is a slight modification of a subset of 802.11i, designed to bring TKIP to the
market more quickly. It is identical to the Robust Security Network information element in Figure 450, but with the following changes:
The element ID is 221, not 48.
A WPA-specific tag of 00:50:F2:01 is inserted before the version field.
Microsoft's OUI (00:50:F2) is used instead of the 802.11 working group's OUI.
Only one cipher suite and one authentication suite are supported in the information element.
However, many WPA implementations do not follow this restriction.
TKIP is the default cipher, rather than CCMP.
Preauthentication is not supported in WPA, so the preauthentication capabilities bit is always
zero.
Types of Management Frames
The fixed fields and information elements are used in the body of management frames to convey
information. Several types of management frames exist and are used for various link-layer
maintenance functions.
Beacon
Beacon frames announce the existence of a network and are an important part of many network
maintenance tasks. They are transmitted at regular intervals to allow mobile stations to find and
identify a network, as well as match parameters for joining the network. In an infrastructure
network, the access point is responsible for transmitting Beacon frames. The area in which Beacon
frames appear defines the basic service area. All communication in an infrastructure network is done
through an access point, so stations on the network must be close enough to hear the Beacons.
Figure 4-51 shows most the fields that can be used in a Beacon frame in the order in which they
appear. Not all of the elements are present in all Beacons. Optional fields are present only when
there is a reason for them to be used. The FH and DS Parameter Sets are used only when the
underlying physical layer is based on frequency hopping or direct-sequence techniques. Only one
physical layer can be in use at any point, so the FH and DS Parameter Sets are mutually exclusive.
The CF Parameter Set is used only in frames generated by access points that support the PCF, which
is optional. The TIM element is used only in Beacons generated by access points, because only access
points perform frame buffering. If the Country-specific frequency hopping extensions were to be
present, they would follow the Country information element. Frequency hopping networks are much
less common now, though, so I omit the frequency hopping extensions for simplicity. Likewise, the
IBSS DFS element occur between the Quiet and TPC Report elements, were it to appear.
Figure 4-51. Beacon frame
Probe Request
Mobile stations use Probe Request frames to scan an area for existing 802.11 networks. The format
of the Probe Request frame is shown in Figure 4-52. All fields are mandatory.
Figure 4-52. Probe Request frame
A Probe Request frame contains two fields: the SSID and the rates supported by the mobile station.
Stations that receive Probe Requests use the information to determine whether the mobile station
can join the network. To make a happy union, the mobile station must support all the data rates
required by the network and must want to join the network identified by the SSID. This may be set
to the SSID of a specific network or set to join any compatible network. Drivers that allow cards to
join any network use the broadcast SSID in Probe Requests.
Probe Response
If a Probe Request encounters a network with compatible parameters, the network sends a Probe
Response frame. The station that sent the last Beacon is responsible for responding to incoming
probes. In infrastructure networks, this station is the access point. In an IBSS, responsibility for
Beacon transmission is distributed. After a station transmits a Beacon, it assumes responsibility for
sending Probe Response frames for the next Beacon interval. The format of the Probe Response
frame is shown in Figure 4-53. Some of the fields in the frame are mutually exclusive; the same rules
apply to Probe Response frames as to Beacon frames.
Figure 4-53. Probe Response frame
The Probe Response frame carries all the parameters in a Beacon frame, which enables mobile
stations to match parameters and join the network. Probe Response frames can safely leave out the
TIM element because stations sending probes are not yet associated and thus would not need to
know which associations have buffered frames waiting at the access point.
IBSS announcement traffic indication map (ATIM)
IBSSs have no access points and therefore cannot rely on access points for buffering. When a station
in an IBSS has buffered frames for a receiver in low-power mode, it sends an ATIM frame during the
delivery period to notify the recipient it has buffered data. See Figure 4-54.
Figure 4-54. ATIM frame
Disassociation and Deauthentication
Disassociation frames are used to end an association relationship, and Deauthentication frames are
used to end an authentication relationship. Both frames include a single fixed field, the Reason Code,
as shown in Figure 4-55. Of course, the Frame Control fields differ because the subtype distinguishes
between the different types of management frames. 802.11 revisions did not need to change the
format, but many have added new reason codes.
Figure 4-55. Disassociation and Deauthentication frames
Association Request
Once mobile stations identify a compatible network and authenticate to it, they may attempt to join
the network by sending an Association Request frame. The format of the Association Request frame
is shown in Figure 4-56.
Figure 4-56. Association Request frame
The Capability Information field is used to indicate the type of network the mobile station wants to
join. Before an access point accepts an association request, it verifies that the Capability Information,
SSID, and (Extended) Supported Rates all match the parameters of the network. Access points also
note the Listen Interval, which describes how often a mobile station listens to Beacon frames to
monitor the TIM. Stations that support spectrum management will have the power and channel
capability information elements, and stations supporting security will have the RSN information
element.
Reassociation Request
Mobile stations moving between basic service areas within the same extended service area need to
reassociate with the network before using the distribution system again. Stations may also need to
reassociate if they leave the coverage area of an access point temporarily and rejoin it later. See
Figure 4-57.
Figure 4-57. Reassociation Request frame
Association and Reassociation Requests differ only in that a Reassociation Request includes the
address of the mobile station's current access point. Including this information allows the new access
point to contact the old access point and transfer the association data. The transfer may include
frames that were buffered at the old access point.
Association Response and Reassociation Response
When mobile stations attempt to associate with an access point, the access point replies with an
Association Response or Reassociation Response frame, shown in Figure 4-58. The two differ only in
the subtype field in the Frame Control field. All fields are mandatory. As part of the response, the
access point assigns an Association ID. How an access point assigns the association ID is
implementation-dependent.
Figure 4-58. (Re)Association Response frame
Authentication
At the beginning of 802.11 networking, stations authenticated using a shared key, and exchanged
Authentication frames, which are shown in Figure 4-59. With 802.11i, shared key authentication was
kept in the standard, but made incompatible with the new security mechanisms. If a station uses
shared key authentication, it will not be allowed to use the strong security protocols described in
Chapter 8.
Figure 4-59. Authentication frames
Different authentication algorithms may co-exist. The Authentication Algorithm Number field is used
for algorithm selection. The authentication process may involve a number of steps (depending on the
algorithm), so there is a sequence number for each frame in the authentication exchange. The Status
Code and Challenge Text are used in different ways by different algorithms; details are discussed in
Chapter 8.
Action frame
802.11h added support for Action frames, which trigger measurements. These frames will be
described in detail in the "Spectrum Management" section of Chapter 8.
Frame Transmission and Association and Authentication
States
Allowed frame types vary with the association and authentication states. Stations are either
authenticated or unauthenticated and can be associated or unassociated. These two variables can be
combined into three allowed states, resulting in the 802.11 Hierarchy of Network Development:
1. Initial state; not authenticated and not associated
2. Authenticated but not yet associated
3. Authenticated and associated
Each state is a successively higher point in the development of an 802.11 connection. All mobile
stations start in State 1, and data can be transmitted through a distribution system only in State 3.
(IBSSs do not have access points or associations and thus only reach Stage 2.) Figure 4-60 is the
overall state diagram for frame transmission in 802.11.
Frame Classes
Frames are also divided into different classes. Class 1 frames can be transmitted in State 1; Class 1
and 2 frames in State 2; and Class 1, 2, and 3 frames in State 3.
Class 1 frames
Class 1 frames may be transmitted in any state and are used to provide the basic operations used by
802.11 stations. Control frames are received and processed to provide basic respect for the CSMA/CA
"rules of the road" and to transmit frames in an IBSS. Class 1 frames also allow stations to find an
infrastructure network and authenticate to it. Table 4-11 shows a list of the frames that belong to the
Class 1 group.
Figure 4-60. Overall 802.11 state diagram
Table 4-11. Class 1 frames
Control
Management
Data
Request to Send
(RTS)
Probe Request
Any frame with ToDS and FromDS
false (0)
Clear to Send (CTS)
Probe Response
Acknowledgment
(ACK)
Beacon
CF-End
Authentication
CF-End+CF-Ack
Deauthentication
Announcement Traffic Indication
Message (ATIM)
Class 2 frames
Class 2 frames can be transmitted only after a station has successfully authenticated to the network,
and they can be used only in States 2 and 3. Class 2 frames manage associations. Successful
association or reassociation requests move a station to State 3; unsuccessful association attempts
cause the station to stay in State 2. When a station receives a Class 2 frame from a nonauthenticated
peer, it responds with a Deauthentication frame, dropping the peer back to State 1.[*] Table 4-12
shows the Class 2 frames.
[*]
This rejection action takes place only for frames that are not filtered. Filtering prevents frames from a different BSS from
triggering a rejection.
Table 4-12. Class 2 frames
Control
Management
Data
None
Association Request/Response
None
Reassociation Request/Response
Disassociation
Class 3 frames
Class 3 frames are used when a station has been successfully authenticated and associated with an
access point. Once a station has reached State 3, it is allowed to use distribution system services and
reach destinations beyond its access point. Stations may also use the power-saving services provided
by access points in State 3 by using the PS-Poll frame. Table 4-13 lists the different types of Class 3
frames.
Table 4-13. Class 3 frames
Control Management
PS-Poll
Data
Deauthentication Any frames, including those with either the ToDS or FromDS bits set
If an access point receives frames from a mobile station that is authenticated but not associated, the
access point responds with a Disassociation frame to bump the mobile station back to State 2. If the
mobile station is not even authenticated, the access point responds with a Deauthentication frame to
force the mobile station back into State 1.
Chapter 5. Wired Equivalent Privacy (WEP)
Anyone who is not shocked by quantum theory has not understood it.
Niels Bohr
In wireless networks, the word "broadcast" takes on an entirely new meaning. Wireless networks rely
on an open medium, and the risk of using them is greatly increased if no cryptographic protection
can be applied on the air link. With an open network medium, unprotected traffic can be seen by
anybody with the right equipment. In the case of wireless LANs, the "right equipment" is a radio
capable of receiving and decoding 802.11, which is hardly an expensive purchase. For extra
eavesdropping power, a high-gain external antenna may be used. Antennas are inexpensive enough
that you must assume that a determined attacher has purchased one.
Guarding against traffic interception is the domain of cryptographic protocols. As frames fly through
the air, they must be protected against harm. Protection takes many forms, but the two most
commonly cited informal objectives are maintaining the secrecy of network data and ensuring it has
not been tampered with. Initially, the Wired Equivalent Privacy (WEP) standard was the answer for
wireless security. In the first four years of 802.11's life, researchers built a strong case for the
insecurity of WEP.
If WEP is so bad, why bother with it? In many cases, it is the only security protocol available on a
particular device. WEP's design is easy to implement. Though it lacks the sophistication of later
cryptographic protocols, it does not require the computational power, either. Older devices, especially
handheld application-specific devices, may lack the processing punch necessary to run anything
better, and WEP is the best that you can do. It is also important to learn about WEP because the WEP
frame-handling operations underlie newer technology such as TKIP.
Cryptographic Background to WEP
Before discussing the design of WEP, it's necessary to cover some basic cryptographic concepts. I am
not a cryptographer, and a detailed discussion of the cryptography involved would not be appropriate
in this book, so this chapter is necessarily brief.[*]
[*]
Readers interested in more detailed explanations of the cryptographic algorithms involved should consult Applied Cryptography
by Bruce Schneier (Wiley, 1996).
To protect data, WEP requires the use of the RC4 cipher, which is a symmetric (secret-key) stream
cipher. RC4 shares a number of properties with all stream ciphers. Generally speaking, a stream
cipher uses a stream of bits, called the keystream. The keystream is then combined with the
message to produce the ciphertext. To recover the original message, the receiver processes the
ciphertext with an identical keystream. RC4 uses the exclusive OR (XOR) operation to combine the
keystream and the ciphertext. Figure 5-1 illustrates the process.
Figure 5-1. Generic stream cipher operation
Most stream ciphers operate by taking a relatively short secret key and expanding it into a
pseudorandom keystream the same length as the message. This process is illustrated in Figure 5-2.
The pseudorandom number generator (PRNG) is a set of rules used to expand the key into a
keystream. To recover the data, both sides must share the same secret key and use the same
algorithm to expand the key into a pseudorandom sequence.
Because the security of a stream cipher rests entirely on the randomness of the keystream, the
design of the key-to-keystream expansion is of the utmost importance. When RC4 was selected by
the 802.11 working group, it appeared to be quite secure. But once RC4 was selected as the
ciphering engine of WEP, it spurred research that ultimately found an exploitable flaw in the RC4
cipher that will be discussed later.
Figure 5-2. Keyed stream cipher operation
Stream Cipher Security
A totally random keystream is called a one-time pad and is the only known encryption scheme that is
mathematically proven to protect against certain types of attacks. One-time pads are not commonly
used because the keystream must be perfectly random and the same length as the data that will be
protected, and it can never be reused.
Attackers are not limited to attacking the underlying cipher. They can choose to exploit any weak
point in a cryptographic system. One famous Western intelligence effort, code-named VENONA,
broke Soviet messages encrypted with one-time pads that were reused. [*] It is easy to understand
the temptation to reuse the one-time pads. Huge volumes of keying material are necessary to protect
even a small amount of data, and those keying pads must be securely distributed, which in practice
proves to be a major challenge. One data bit must have a corresponding one-time pad bit. A 54 Mbps
network can move about 25 Mbps of user data. If the network operates at just 10% of that capacity,
the data transfer in an 8-hour workday is still 9 gigabytes. Distributing multiple gigabytes of keying
material to every access point is totally impractical.
[*]
The United States National Security Agency has made some information on the project public at
http://www.nsa.gov/docs/venona.
Stream ciphers are a compromise between security and practicality. The perfect randomness (and
perfect security) of a one-time pad is attractive, but the practical difficulties and cost incurred in
generating and distributing the keying material is worthwhile only for short messages that require
the utmost security. Stream ciphers use a less random keystream but one that is random enough for
most applications.
Cryptographic Politics
No discussion of cryptography would be complete without a passing reference to some of the many
legal and regulatory concerns surrounding its use. Three major issues impinge upon the use of WEP,
though the effect of these issues has diminished over time.
WEP requires the use of the RC4 cipher to encrypt the frame. When the first edition of this book was
written, WEP was optional and not incorporated into all products. Host software could incorporate
RC4 code, but open source projects had concerns about including code that infringed on the
intellectual property of RSA Security, Inc. In the time since the first edition was published, this
concern has faded into the background. All the major chipset vendors have licensed RC4 encryption
and incorporated hardware support for RC4 into 802.11 chipsets. Device drivers are responsible for
pushing WEP keys down to the hardware. Performing encryption in hardware on the card means that
software no longer needs to risk infringing on RSA's intellectual property.
WEP was initially designed with short keys to satisfy the U.S. export regulations regarding
cryptographic products. Initially, the standard required short 40-bit keys, but every product on the
market supports at least 104-bit keys now. For a brief time, long keys looked like an important
extension to the standard, though the additional security proved illusory.
Finally, some governments strictly regulate the use of any cryptographic system, WEP included. In
addition to United States export regulations, many countries have import regulations that restrict
importing cryptographic equipment. Other governments are also free to require additional
cryptographic measures. The government of China has developed an alternative security system
called WLAN Authentication and Privacy Infrastructure (WAPI), and has made it optional for wireless
LAN equipment sold in China.
WEP Cryptographic Operations
Communications security has three major objectives. Any protocol that attempts to secure data as it
travels across a network must help network managers to achieve these goals. Confidentiality is the
term used to describe data that is protected against interception by unauthorized parties. Integrity
means that the data has not been modified. Authentication underpins any security strategy because
part of the reliability of data is based on its origin. Users must ensure that data comes from the
source it purports to come from. Systems must use authentication to protect data appropriately.
Authorization and access control are both implemented on top of authentication. Before granting
access to a piece of data, systems must find out who the user is (authentication) and whether the
access operation is allowed (authorization).
WEP provides operations that attempt to meet these objectives, though they often fail under serious
scrutiny or attack. Frame body encryption supports confidentiality. An integrity check sequence
protects data in transit and allows receivers to validate that the received data was not altered in
transit. In practice, the operations offered by WEP are not strong and require additional
strengthening measures, as discussed in the next two chapters.
WEP Data Processing
Confidentiality and integrity are handled simultaneously, as illustrated in Figure 5-3. Before
encryption, the frame is run through an integrity check algorithm, generating a hash called an
integrity check value (ICV). The ICV protects the contents against tampering by ensuring that the
frame has not changed in transit. The frame and the ICV are both encrypted, so the ICV is not
available to casual attackers.
Figure 5-3. WEP operations
As input, WEP requires three items:
The payload to be protected, which comes from the upper layer protocol stack.
A secret key, used in frame encryption. Depending on implementation, keys may be specified as
a string of key bits, or by key number. WEP allows four keys to be stored simultaneously.
An initialization vector, used along with the secret key in frame transmission.
After processing, WEP has a single output:
An encrypted frame, ready for transmission over an untrusted network with enough information
to enable decryption at the remote end.
WEP data transmission
Driver software and the interface hardware are responsible for processing the data and sending out
the encrypted packet, using the following sequence:
1. The 802.11 frame is queued for transmission. It consists of a frame header and the payload.
WEP protects only the payload of the 802.11 MAC, and leaves the 802.11 frame header, as well
as any lower-layer headers, intact.
2.
1.
2. An integrity check value (ICV) is calculated over the payload of the 802.11 MAC frame. It is
calculated over the frame payload, so it starts at the first bit of the SNAP header, and goes up
to the last data bit in the body. The 802.11 frame check sequence has not yet been calculated,
so it is not included in the ICV calculation. The ICV used by WEP is a Cyclic Redundancy Check
(CRC), a point that will be expanded on later.
3. The frame encryption key, or WEP seed, is assembled. WEP keys come in two parts: the secret
key, and the initialization vector (IV). Stream ciphers will produce the same key stream from
the same key, so an initialization vector is used to produce different stream ciphers for each
transmitted frame. To reduce the occurrence of encryption with the same key stream, the
sending station prepends the IV to the secret key. 802.11 does not place any constraints on the
algorithm used to choose IVs; some products assign IVs sequentially, while others use a
pseudorandom hashing algorithm. IV selection has some security implications because poor IV
selection can compromise keys.
4. The frame encryption key is used as the RC4 key to encrypt the 802.11 MAC payload from step
1 and the ICV from step 2. The encryption process is often assisted with dedicated RC4 circuitry
on the card.
5. With the encrypted payload in hand, the station assembles the final frame for transmission. The
802.11 header is retained intact. Between the 802.11 MAC header and the encrypted payload, a
WEP header is inserted. In addition to the IV, the WEP header includes a key number. WEP
allows up to four keys to be defined, so the sender must identify which key is in use. Once the
final header is assembled, the 802.11 FCS value can be calculated over the entire MAC frame
from the start of the header to the end of the (encrypted) ICV.
Decryption happens in the reverse order. As with any other transmission on the wireless network, the
FCS is validated to ensure that the frame received was not corrupted in transit. To decrypt the
protected part of the frame, the receiver will take its secret key, prepend the IV, and generate the
key stream. With the decrypted data, it can then validate the ICV. Once the ICV is successfully
validated, the packet data can be passed to the appropriate upper-layer protocol depending on the
content of the SNAP header.
WEP key length
In theory, WEP can be used with keys of any length because RC4 does not require the use of any
particular key size. Most products implement one of two key lengths, though. The only key length
present in the standard is a 64-bit WEP seed, of which 40 bits are shared as a secret between the
two communicating stations. Vendors have used a variety of names for the standard WEP mode:
"standard WEP," "802.11-compliant WEP," "40-bit WEP," "40+24-bit WEP," or even "64-bit WEP." I
personally feel that the last term is a stretch, based on hoodwinking the consumer with the length of
the shared key and not the size of the shared secret, but it has become somewhat standard
throughout the industry.
It is also common to implement a longer key length, typically with a 128-bit WEP seed, of which 104
bits are a kept secret. This is referred to as "WEP-104" by many documents, though it is almost
always called "128-bit WEP" by sales literature. It is rare, though not unheard of, to use a third key
length of 128 secret bits, which results in a 152-bit WEP seed. Confusingly, this rare implementation
is often also called "128-bit WEP," although the differing key lengths makes it incompatible with
WEP-104. One vendor even offers the option of using 256-bit secret keys with WEP, although the
incremental security from such a long key is dubious.
In a well-designed cryptographic system, additional security can be obtained by using a longer key.
Each additional bit doubles the number of potential keys and, in theory, doubles the amount of time
required for a successful attack. WEP, however, is not a well-designed cryptographic system, and the
extra bits in the key buy you very little. The best publicly disclosed attack against static WEP can
recover the key in seconds, no matter what its length is.
Types of WEP keys
WEP can employ two different types of keys. Mapped keys protect traffic between a particular source
and receiver. Mapped keys are sometimes referred to as unicast keys or station keys because they
are well-suited to protect unicast traffic. In an infrastructure network, data is always transferred
between a station and the access point serving it. 802.11 frames have both a receiver address and a
destination address. Unicast traffic may have many different destination addresses, but every unicast
frame transmitted in an infrastructure network will use the MAC address of the AP as its receiver
address. All unicast traffic to or from a particular station can be encrypted by a single mapping key.
Default keys, which are also referred to as broadcast keys, are used when no mapping relationship
exists between two 802.11 stations. Default keys are a natural fit for broadcast and multicast frames
because group addresses represent multiple stations and therefore cannot support key mapping
relationships.
Manual (static) versus automatic (dynamic) WEP
802.11 does not specify any particular key distribution mechanism for use with WEP. As far as WEP is
concerned, keys magically appear. Early WEP implementations relied on manual key distribution.
Administrators were responsible for distributing a single default key to all stations in the network,
usually manually. Key updates were also manual. In practice, most network deployments wound up
using the same key for a very long time due to the high management overhead in changing the key.
For this reason, WEP without any key distribution mechanism is often called manual WEP or static
WEP.
Manual keys are an awful idea. Managing key distribution across a whole network of access points is
very difficult. Keys protect data, and should therefore be changed whenever somebody with
knowledge of the key leaves the organization. In theory, a well designed cryptographic system tries
to limit the use of a key to as narrow a purpose as possible just in case a key is compromised. Static
WEP used the same key for all frames transmitted by every station, which worsens the consequences
of a key compromise.
Static WEP should not be used without a clear reason to do so. Static WEP is better than nothing, but
little more can be said in its favor. It is effective for deterring casual attacks, but will not withstand
determined assault. Many low-power devices such as 802.11 phones, handheld bar code scanners,
and even some PDAs may not support anything better than static WEP. Until they can be replaced,
static WEP offers the only possible link layer security for these devices.
Better solutions are based on dynamic WEP. Rather than a single key distributed to all stations, each
station uses two keys. One is a key mapping key, shared between the station and access point, used
to protect unicast frames. The second key is a default key, shared by all stations in the same service
set, that protects broadcast and multicast frames. The encryption keys used by the clients are
distributed using key encryption keys derived from strong authentication protocols that will be
discussed in the next chapter.
Dynamic WEP offers significant advantages over static WEP solutions. At the highest level, the scope
of each key is limited. Keys are used less often, and the result of compromising a key is lessened by
using it to protect less traffic. Attackers can still mount assaults on the keys, but they have much less
data to work for each key, making attacks more time-consuming. Nearly as important, dynamic keys
can, as the name implies, change. At periodic intervals, the keys can be refreshed by the access
point, which requires attackers to throw out accumulated data and start the assault on a key over
again. By configuring short rekeying intervals, significant mitigation against attacks on keys is
possible. Dynamic WEP solutions will be discussed later in this chapter.
WEP key numbering and storage
WEP keys have an associated number so that up to four keys may be defined on an 802.11 station.
In the beginning, the WEP key number was meant to help facilitate changing the key across the
network. Rather than switch every station to a new key simultaneously, the new key could be defined
across the network and stations could be gradually switched. If, for example, an organization used
key number 0 and wished to change it, key number 1 could be defined on the network. Rather than
change all stations at the same time to the new key, stations could be configured to start using key
number 1 during a defined transition period. At the end of the transition period, key number 0 would
be disabled.
Key numbering takes on a slightly different role in dynamic WEP solutions. Each station receives two
keys from the access point: a mapping key, typically stored as key number 0, and a default key,
typically stored as key number 1. Stations use key 0 for protection of unicast traffic, and key 1 for
protection of broadcast traffic. When frames are queued by the driver for transmission, the driver
software will label unicast frames as requiring encryption with key 0, and broadcast frames as
requiring encryption with key 1.
When 802.11 was new, many cards did not support WEP, or performed the RC4 encryption on the
host CPU, limiting performance. RC4 is very easy to implement in hardware, and virtually every
wireless LAN interface sold since 1999 has provided hardware-based RC4 encryption in the interface
card itself to reduce the performance hit from encryption. (This includes most 802.11b cards and
everything that has come later.) Some older cards also used firmware to limit key length to 40-bit
keys rather than 104-bit keys, such as the Orinoco Silver and Gold cards. Key length restrictions
were lifted in later firmware releases.
To efficiently encrypt frames, many 802.11 chipsets include a data structure called the key cache.
Key caches consist of mappings between tuples of the destination address, the key identifier number,
and the bits of the key itself. Most chipsets intended for use in station interface cards have four key
slots. Static WEP uses one key slot; dynamic WEP solutions use two. When frames are queued for
transmission, the destination address is looked up in the key cache and the resulting key is used to
encrypt the frame. Dynamic keying solutions function identically to static WEP solutions, except that
the contents of the key cache may be overwritten by key management software.
One of the ways in which chipsets intended for use in cards differ from chipsets intended for use in
access points is that the AP chipsets have much larger key caches, often capable of 256 or more key
entries. Even with two entries required (one mapped key and one default key) for a station, such a
data structure can handle keying for more than 100 stations. Given the limited capacity of most
wireless link technologies, such a large data will handle any reasonable traffic load with room to
spare. However, some early APs capable of dynamic keying were unable to maintain large key data
structures, and had to share keys among multiple stations. To cut costs, some vendors used card
chipsets rather than AP chipsets in the AP radio. As a result, they only had four key slots and needed
to share a single unicast key among all stations attached to an AP.
WEP Encapsulation
When WEP is in use, the frame body expands by eight bytes. Four bytes are used for a frame body IV
header, and four are used for the ICV trailer. See Figure 5-4.
Figure 5-4. WEP frame extensions
The IV header uses 3 bytes for the 24-bit IV, with the fourth byte used for padding and key
identification. When a default key is used, the Key ID subfield identifies the default key that was used
to encrypt the frame. If a key mapping relationship is used, the Key ID subfield is 0. The 6 padding
bits of the last byte must be 0. The integrity check is a 32-bit CRC of the data frame; it is appended
to the frame body and protected by RC4. The frame check sequence protects the encrypted data.
Problems with WEP
Cryptographers have identified many flaws in WEP. The designers specified the use of RC4, which is
widely accepted as a strong cryptographic cipher. Attackers, however, are not limited to a full-frontal
assault on the cryptographic algorithmsthey can attack any weak point in the cryptographic system.
Methods of defeating WEP have come from every angle. One notable slip was a vendor who shipped
access points that exposed the secret WEP keys through SNMP, allowing an attacker to ask for just
the key. Most of the press, though, has been devoted to flaws beyond implementation errors, which
are much harder to correct.
Cryptographic Properties of RC4
Reuse of the keystream is the major weakness in any stream cipher-based cryptosystem. When
frames are encrypted with the same RC4 keystream, the XOR of the two encrypted packets is
equivalent to the XOR of the two plaintext packets. By analyzing differences between the two
streams in conjunction with the structure of the frame body, attackers can learn about the contents
of the plaintext frames themselves. To help prevent the reuse of the keystream, WEP uses the IV to
encrypt different packets with different RC4 keys. However, the IV is part of the packet header and is
not encrypted, so eavesdroppers are tipped off to packets that are encrypted with the same RC4 key.
Implementation problems can contribute to the lack of security. 802.11 admits that using the same
IV for a large number of frames is insecure and should be avoided. The standard allows for using a
different IV for each frame, but it is not required.
WEP incorporates an integrity check, but the algorithm used is a cyclic redundancy check (CRC).
CRCs can catch single-bit alterations with high probability, but they are not cryptographically secure.
Cryptographically secure integrity checks are based on hash functions, which are unpredictable. With
unpredictable functions, if the attacker changes even one bit of the frame, the integrity check will
change in unpredictable ways. The odds of an attacker finding an altered frame with the same
integrity check are so slim that it cannot be done in real time. CRCs are not cryptographically secure.
CRC calculations are straightforward mathematics, and it is easy to predict how changing a single bit
will affect the result of the CRC calculation. (This property is often used by compressed data files for
repair! If just a few bits are bad, they can sometimes be identified and corrected based on a CRC
value.)
Design Flaws of the WEP System
WEP's design flaws initially gained prominence when the Internet Security, Applications,
Authentication, and Cryptography (ISAAC) group at the University of California, Berkeley, published
preliminary results based on their analysis of the WEP standard.[*] None of the problems identified
by researchers depend on breaking RC4. Here's a summary of the problems they found; I've already
touched on some of them:
[*]
The report is available on the Web at http://www.isaac.cs.berkeley.edu/isaac/wep-faq.html. Items 36 on the following list are
summarized from that report.
1.
1. Manual key management is a minefield of problems. Setting aside the operational issues with
distributing identical shared secrets to the user population, the security concerns are
nightmarish. New keying material must be distributed on a "flag day" to all systems
simultaneously, and prudent security practices would lean strongly toward rekeying whenever
anybody using WEP leaves the company (the administrative burden may, however, preclude
doing this). Widely distributed secrets tend to become public over time. Passive sniffing attacks
require obtaining only the WEP keys, which are likely to be changed infrequently. Once a user
has obtained the WEP keys, sniffing attacks are easy, especially since 802.11 analyzers
incorporate decryption capabilities.
2. As standardized, static WEP offers a shared secret of only 40 bits. Security experts have long
questioned the adequacy of 40-bit private keys, and many recommend that sensitive data be
protected by at least 128-bit keys.[*] In the years since the initial flaws in WEP were published,
the industry-standard extended key length is only 104 bits.
[*]
To be fair, WEP was originally developed with the goal of being exportable under the then-current U.S.
regulations for export of cryptographic systems. A longer key could not have been used without
jeopardizing the commercial viability of U.S.-built 802.11 products.
3. Stream ciphers are vulnerable to analysis when the keystream is reused. WEP's use of the IV
tips off an attacker to the reuse of a keystream. Two frames that share the same IV almost
certainly use the same secret key and keystream. This problem is made worse by poor
implementations, which may not pick random IVs. The Berkeley team identified one
implementation that started with an IV of 0 when the card was inserted and simply incremented
the IV for each frame. Furthermore, the IV space is quite small (less than 17 million), so
repetitions are guaranteed on busy networks.
4. Infrequent rekeying allows attackers to assemble what the Berkeley team calls decryption
dictionaries large collections of frames encrypted with the same keystreams. As more frames
with the same IV pile up, more information is available about the unencrypted frames even if
the secret key is not recovered. Given how overworked the typical system and network
administration staff is, infrequent rekeying is the rule.
5. WEP uses a CRC for the integrity check. Although the value of the integrity check is encrypted
by the RC4 keystream, CRCs are not cryptographically secure. Use of a weak integrity check
does not prevent determined attackers from transparently modifying frames.[ ]
[
] 802.11 requires frame retransmissions in the case of loss, so it may be possible for an attacker to
retransmit a frame and cause a replacement injected frame to be accepted as legitimate.
6. The access point is in a privileged position to decrypt frames. Conceptually, this feature can be
attacked by tricking the access point into retransmitting frames that were encrypted by WEP.
Frames received by the access point would be decrypted and then retransmitted to the
attacker's station. If the attacker is using WEP, the access point would helpfully encrypt the
frame using the attacker's key.
Key Recovery Attacks Against WEP
In August 2001, Scott Fluhrer, Itsik Mantin, and Adi Shamir published a paper titled "Weaknesses in
the Key Scheduling Algorithm of RC4." At the end of the paper, the authors describe a theoretical
attack on WEP. At the heart of the attack is a weakness in the way that RC4 expands the key into the
key stream, a process called the key scheduling algorithm.
The Fluhrer-Mantin-Shamir attack, or FMS attack for short, assumes the ability to recover the first
byte of the encrypted payload. In many cryptographic systems, that would be a big assumption.
WEP, however, is used to protect 802.11 frames, and the 802.11 frame body begins with a SNAP
header. Therefore, the cleartext value of the first byte is known to be 0xAA. Because the first
cleartext byte is known, the first byte of the keystream can be easily deduced from a trivial XOR
operation with the first encrypted byte.
The paper's attacks are focused on a class of weak keys written in the form (B+3):FF:N. Each weak
IV is used to attack a particular byte of the secret portion of the RC4 key. Key bytes are numbered
from zero. Therefore, the weak IV corresponding to byte zero of the secret key has the form 3:FF:N.
The second byte must be 0xFF; knowledge of the third byte in the key is required, but it need not be
any specific value.
A standard WEP key is 40 secret bits, or 5 bytes numbered consecutively from 0 to 4. Weak IVs on a
network protected by standard WEP must have a first byte that ranges from 3 (B=0) to 7 (B=4) and
a second byte of 255. The third byte must be noted but is not constrained to any specific value.
There are 5 x 1 x 256=1,280 weak IVs of this type in a static WEP network. Additional classes of
weak IVs have been exploited since the publication of the paper, bringing the total number of weak
IVs to about 9,000; this is approximately 5% of the total number of IVs. These new classes depend
on the relationships between IV bytes. Readers who are interested in the other classes of weak IVs
should refer to the source code of WEP cracking tools. [*]
[*]
For example, see the classify() function in Airsnort's crack.c.
Each weak IV leaks information about a key byte. By applying probability theory, Fluhrer, Mantin,
and Shamir estimated that about 60 weak IVs need to be collected for each key byte. Furthermore,
and perhaps worst of all, the attack gains speed as more key bytes are determined. Guessing the
first key byte helps you get the second, and so on. Success cascades through the attack, and it
works in linear time. Doubling the key length only doubles the computational time for the attack to
succeed.
With such a tantalizing result, it was only a matter of time before it was used to attack a real system.
In early August 2001, Adam Stubblefield, John Ioannidis, and Avi Rubin applied the FMS attack to an
experimental, but real, network with devastating effect.[*] In their testing, 60 resolved cases usually
determined a key byte, and 256 resolved cases always yielded a full key. It took less than a week to
implement the attack, from the ordering of the wireless card to the recovery of the first full key.
Coding the attack took only a few hours. Key recovery was accomplished between five and six million
packets, which is a small number for even a moderately busy network.
[*]
This work is described more fully in AT&T Labs Technical Report TD-4ZCPZZ.
Reporting on a successful attack, however, is nothing compared to having a public code base
available to use at will. The hard part of the Fluhrer-Mantin-Shamir attack was finding the RC4
weakness. Implementing their recommendations is not too difficult. In late August 2001, AirSnort, an
open source WEP key recovery program, was released. Many tools have followed AirSnort.
Key recovery defenses
Longer keys are no defense against key recovery attacks. The time required to recover a key can be
broken up into the gathering time required to collect enough frames for the attack, plus the
computational time required to run the program and get the key. Computational time is only a few
seconds; gathering time generally dominates the attack. Longer keys require slightly more
computational time, but do not require different gathering time. As the key length increases, more
weak IVs are caught in the dragnet.
One defense adopted by many vendors is to avoid using weak IVs. Most vendors have now changed
products so that each IV to be used is first checked against a classifier, and any weak IVs are
replaced by non-weak IVs. Unfortunately, reducing the size of the IV space may cause IV re-use to
happen earlier.
Network administrators have responded to key recovery attacks by using stronger protocols, such as
the 802.11i protocols discussed in Chapter 7. For many organizations, this is a solution. However,
newer protocols may not be as compatible as dynamic WEP with existing equipment. To mitigate the
use of WEP, most network administrators set a re-key time of between 5 and 15 minutes, depending
on the acceptable overhead on the network.
Dynamic WEP
With the attention devoted to security, it was not long before the industry started to develop wireless
LAN technologies with significantly improved security. The first step along this road was to bolster
WEP by refreshing the keys dynamically. Rather than a single static WEP key for all frames on the
network shared by all stations, dynamic WEP solutions use a set of different keys. All stations in a
network share a key to encrypt broadcast frames, and each station has its own mapping key for
unicast frames.[*]
[*]
With appropriate key distribution protocols in place, you can even have multiple groups share the same infrastructure. Each
station has its own mapping key for unicast frames, but there may be multiple independent broadcast groups, each with its own
default key for group frames.
WEP did not specify a key management framework. Keys are generated and distributed through a
system that is not written down in 802.11. The first, and easiest key management "framework" to be
implemented was manual. Network administrators needed to come up with a string of bits used as a
key, and then distribute that key to all stations participating in an 802.11 service set secured by
WEP. To set a key, administrators needed to touch a machine to configure a new key.
Layered Security Protocols
When the first edition of this book was written, WEP was rightly viewed as an insecure
security system. Classic old-school manually-distributed static-key WEP is fundamentally
broken. When it became clear that the relatively minimal protection provided by WEP
was insufficient for most network environments, a task group formed to develop
enhancements to the security of the MAC. That work was standardized as 802.11i in June
of 2004.
During the gap between the initial research that exposed WEP's flaws and the
development of more secure technologies to bolster it, network administrators turned to
proven security protocols at higher layers in the stack, such as IPsec (layer 3), SSL
(layer 4), and SSH (layer 7). With static WEP offering only minimal security, the
additional encryption strength provided by higher-layer technology was well worth it.
With the development of stronger link layer technologies, layered security protocols are
no longer the magic bullet they once appeared to be. IPsec requires client software to be
installed and configured. SSL-based VPNs are much simpler to set up, though they often
have shortcomings when securing applications that are not web-based. SSH is wellunderstood and can create arbitrary TCP tunnels, but it often requires significant
modifications to user procedures. (Chances are that if you are reading this book, SSH is
second nature; however, it is probably not something you want your users trying to
figure out.)
With the development of the improved link layer security technologies discussed in this
chapter and the next two, it is finally possible to build secure networks at the link layer.
One increasingly common approach to building secure wireless networks is to consider
what these new technologies can do, and then determine whether additional protection
at higher layer protocols is required. Balancing the trade-offs is the subject of Chapter
22, in the deployment section of the book.
Dynamic WEP uses an improved key management framework. Rather than depend on the
administrator for so much manual work, dynamic WEP uses strong cryptographic protocols to
generate keys and then distribute them, in encrypted form, over untrusted networks. WEP key
generation typically depends on the use of a cryptographic authentication protocol, which is
discussed in the next chapter.
Dynamic WEP handles frames in an identical fashion to static WEP. The only difference is that there is
a much improved mechanism to generate and distribute keys on a periodic basis. The automatic key
management of dynamic WEP achieves much greater security than static WEP because it dramatically
shortens the lifetime of a key. Any attacks against the key must take place within a single key
lifetime. Frame initialization vectors can be re-used after a key refresh because they correspond to
two different WEP seeds. Key recovery attacks using Fluhrer/Mantin/Shamir must occur within a
single key lifetime as well, for obvious reasons. Dynamic WEP is by no means perfect, but it is a
substantial improvement over static WEP. It is widely supported by almost every card and driver.
Chapter 6. User Authentication with 802.1X
What is your name?
What is your quest?
What is your favorite color?
The Bridgekeeper
Monty Python and the Holy Grail
Security is a common thread linking many of the wireless LAN stories in the news throughout the
past several years, and polls repeatedly show that network managers consider security to be a
significant obstacle to wider deployment of wireless LANs. Many of the security problems that have
prevented stronger acceptance of 802.11 are caused by flaws in the design of static WEP.
Manual WEP attempts to be too many solutions to multiple problems. It was intended to be used both
for authentication, by restricting access to those in possession of a key, and confidentiality, by
encrypting data as it traversed wireless links. In the final analysis, it does neither particularly well.
Both authentication and confidentiality are important issues for wireless LANs, and the subject of a
great deal of technology development since the first edition of this book.
This chapter takes on the problem of authentication, which is provided at the link layer through the
use of 802.1X.[*] 802.1X has matured a great deal since the first edition of this book, and is
increasingly the authentication protocol of choice on wireless LANs.[ ] Static WEP authenticates
machines in possession of a cryptographic key. 802.1X allows network administrators to authenticate
users rather than machines, and can be used to ensure that users connect to legitimate, authorized
networks rather than credential-stealing impostor networks.
[*]
802.1X is written with a capital X. In the IEEE nomenclature, lowercase letters ("802.11a" and "802.11b") are reserved for addon specifications that revise an existing standard. Uppercase letters are used for standalone specifications. As a full, standalone
protocol specification in its own right, 802.1X gets a capital letter.
[
] One of my personal yardsticks for the maturity of a specification is the existence of an open source implementation. Open
source software frequently serves a valuable role by keeping proprietary implementations honest, and providing a low-cost reality
check for users. In the 802.1X world, the xsupplicant and wpa_supplicant projects have taken on this role.
Identifying users instead of machines can lead to more effective network architecture. Rather than
grouping users by function and applying security controls to the physical ports in a physical location,
the identity of the user and any access rights can be integrated into the network switch fabric, and
follow users around the network. Wireless LANs are often the first use of identity-based policy
enforcement. It is not uncommon for companies to use the capability on wireless networks, and then
find it so useful that it is later integrated into the wired network. No matter where or how users
attach to the network, policy follows them around.
One of the complexities in dealing with 802.1X is that it is a framework. It is an IEEE adaptation of
the IETF's Extensible Authentication Protocol (EAP), originally specified in RFC 2284 and updated by
RFC 3748. EAP is a framework protocol. Rather than specifying how to authenticate users, EAP allows
protocol designers to build their own EAP methods, subprotocols that perform the authentication
transaction. EAP methods can have different goals, and therefore, often use many different methods
for authenticating users depending on the requirements of a particular situation. Before a detailed
discussion of how the different methods work, though, a detailed understanding of how EAP works is
necessary.
The Extensible Authentication Protocol
802.1X is based on EAP. Recent work in wireless networking required an update to the standard,
which is now found in RFC 3748. Back in the early 1990s, when PPP was first introduced, there were
two protocols available to authenticate users, each of which required the use of a PPP protocol
number. Authentication is not a "one size fits all" problem, and it was an active area of research at
the time. Rather than burn up PPP protocol numbers for authentication protocols that might become
obsolete, the IETF standardized EAP. EAP used a single PPP protocol number while supporting a wide
variety of authentication mechanisms. EAP is a simple encapsulation that can run over any link layer,
but it has been most widely deployed on PPP links. Figure 6-1 shows the basic EAP architecture,
which is designed to run over any link layer and use any number of authentication methods.
Figure 6-1. EAP architecture
EAP Packet Format
Figure 6-2 shows the format of an EAP packet. When used on PPP links, EAP is carried in PPP frames
with a protocol number of 0xC227. There is no strict requirement that EAP run on PPP; the packet
shown in Figure 6-2 can be carried in any type of frame. The fields in an EAP packet are:
Code
The Code field, the first field in the packet, is one byte long and identifies the type of EAP
packet. It is used to interpret the Data field of the packet.
Identifier
The Identifier field is one byte long. It contains an unsigned integer used to match requests
with responses to them. Retransmissions reuse the same identifier numbers, but new
transmissions use new identifier numbers.
Length
The Length field is two bytes long. It is the number of bytes in the entire packet, which
includes the Code, Identifier, Length, and Data fields. On some link layer protocols, padding
may be required. EAP assumes that any data in excess of the Length field is link-layer padding
and can be ignored.
Data
The last field is the variable-length Data field. Depending on the type of packet, the Data field
may be zero bytes long. Interpretation of the Data field is based on the value of the Code field.
Figure 6-2. EAP packet format
EAP Requests and Responses
EAP exchanges are composed of requests and responses. The authenticator sends requests to the
system seeking access, and based on the responses, access may be granted or denied. Client
systems only send response packets when there is an outstanding request to send. There is no such
thing as an unsolicited packet from the system seeking authentication. The format of request and
response packets is shown in Figure 6-3.
Figure 6-3. EAP Request and EAP Response packets
The Code field is set to 1 for requests and 2 for responses. The Identifier and Length fields are used
as described in the previous section on the generic format. The Data field carries the data used in
requests and responses. Each Data field carries one type of data, broken down into a type identifier
code and the associated data:
Type
The Type field is a one-byte field that indicates the type of request or response. Only one type
is used in each packet. With one exception, the Type field of the response matches the
corresponding request. That exception is that when a request is unacceptable, the peer may
send a NAK to suggest an alternative type. Types greater than or equal to 4 indicate
authentication methods.
Type-Data
The Type-Data field is a variable field that must be interpreted according to the rules for each
type, as described in the following sections.
Type code 1: Identity
The authenticator generally uses the Identity type as the initial request, which is often written as
EAP-Request/Identity, or simply Request/Identity to indicate that the authenticator is attempting to
establish some sort of username to authenticate. Frequently, the EAP Identity is the user identifier,
possibly with routing information. Some technologies work by submitting an EAP Identity that may
correspond to the machine. In the initial Request/Identity packet, if any information is present in the
Type-Data field, it is used to prompt the user, though this is relatively uncommon. If the Type-Data
field is present and contains a prompt string, its length is inferred from the length of the EAP packet,
rather than having a separate delimiter.
Many EAP implementations are capable of prompting the (human) user for input to determine the
user identity, though it is not required. For greater usability, most EAP implementations also allow
the identity to be statically configured. Once the user name has been determined, the EAP client will
respond with a Response/Identity packet. In Response/Identity packets, the Type-Data field contains
the username. It may be a "bare" username, such as mgast, or it may be qualified with an Internetstyle domain (mgast@domain.com), or a Windows-style domain name (DOMAIN\mgast).
Some EAP implementations may attempt to look up the user identity in a Response even before
issuing the authentication challenge. If the user does not exist, the authentication can fail without
further processing. Most implementations automatically reissue the identity request to give the user
an opportunity to correct typos.
Type code 2: Notification
The authenticator can use the Notification type to send a message to the user. The user's system can
then display the message in the Request/Notification for the user's benefit. Notification messages are
used to provide messages to the user from the authentication system, such as a password about to
expire, or the reason for an account lockout. Notification messages are not commonly used with
802.1X; only a few vendors implement them. Responses must be sent in reply to Notification
requests. However, Response/Notification packets serve as simple acknowledgments, and the TypeData field has a zero length.
Type code 3: NAK
Null acknowledgments (NAKs) are used to suggest a new authentication method. The authenticator
issues a challenge, encoded by a type code. Authentication types are numbered 4 and above. If the
end user system does not support the authentication type of the challenge, it can issue a NAK. The
Type-Data field of a NAK message includes a single byte corresponding to the suggested
authentication type. Most 802.1X implementations do not actively negotiate, and will simply log an
error message if the client attempts to use an unsupported type.
EAP Authentication Methods
In addition to flow-control messages and negotiation messages, EAP assigns type codes to
authentication methods. EAP delegates the work of proving a user identity to a subsidary protocol,
the EAP method, which is a set of rules for authenticating a user.
The advantage of using the method construction is that it frees EAP from any particular set of
assumptions about what is necessary to authenticate a user. When requirements change, as they did
with the popularity of wireless networks, new EAP methods can be developed to meet the challenge.
Table 6-1 lists several EAP methods with their type codes. A more detailed description of the EAP
methods commonly used on wireless LANs will follow later in this chapter.
Table 6-1. Common EAP methods for 802.1X authentication
Type
code
Authentication
protocol
Description
4
MD5 Challenge
CHAP-like authentication in EAP
6
GTC
Originally intended for use with token cards such as RSA
SecurID
13
EAP-TLS
Mutual authentication with digital certificates
21
TTLS
Tunneled TLS; protects weaker authentication methods with TLS
encryption
25
PEAP
Protected EAP; protects weaker EAP methods with TLS
encryption
18
EAP-SIM
Authentication by mobile phone Subscriber Identity Module
(SIM)
29
MS-CHAP-V2
Microsoft encrypted password authentication; compatible with
Windows domains
EAP Success and Failure
At the conclusion of an EAP exchange, the user has either authenticated successfully or has failed to
authenticate (Figure 6-4). Once the authenticator determines that the exchange is complete, it can
issue an EAP-Success (code 3) or EAP-Failure (code 4) frame to end the EAP exchange.
Implementations are allowed to send multiple requests before failing the authentication to allow a
user to get the correct authentication data.
Figure 6-4. EAP Success and Failure frames
Success and Failure frames are not authenticated in any way. In the dial-up world, the telephone
network provides a modicum of security that the sender is at the other end of the circuit. In wireless
LANs, the lack of authentication on Success and Failure frames may require extra protocol design.
A Sample EAP Exchange
A sample EAP exchange is shown in Figure 6-5. It is not a "real" exchange that would be seen on a
wireless network because it uses protocols that are not in wide deployment. It is intended only to
give you a basic idea of how the protocol is supposed to work. The EAP exchange is a series of steps
beginning with a request for identity and ending with a success or failure message. As a matter of
notation, packets transmitted as part of an EAP method exchange are written Request/Method when
they come from the authenticator, and Response/Method when they are sent in response.
1. The authenticator issues a Request/Identity packet to identify the user. Request/Identity
packets serve two purposes. In addition to starting the exchange, they also serve notice to the
client that the network is likely to drop any data traffic before authentication completes.
2. The end user system prompts for input, collects the user identifier, and sends the user identifier
in a Response/Identity message.
3. With the user identified, the authenticator can issue authentication challenges. In step 3 in the
figure, the authenticator issues an MD-5 Challenge to the user with a Request/MD-5 Challenge
packet.
4. The user system is configured to use a token card for authentication, so it replies with a
Response/NAK, suggesting the use of Generic Token Card authentication.
5. The authenticator issues a Request/Generic Token Card challenge, prompting for the numerical
sequence on the card.
6. The user types a response, which is passed along in a Response/Generic Token Card.
7. If the user response is not correct, authentication is not possible. However, the authenticator
EAP implementation allows for multiple authentication Requests, so a second Request/Generic
8.
6.
7.
Token Card is issued.
8. Once again, the user types a response, which is passed along in a Response/Generic Token
Card.
9. On the second try, the response is correct, so the authenticator issues a Success message.
Figure 6-5. Sample EAP exchange
EAP Methods
The E in EAP is both its greatest strength and greatest weakness. Extensibility allows a protocol to
develop new features as new requirements present themselves. EAP has gone from a way to
preserve PPP protocol numbers into the cornerstone of wireless LAN security because of its
extensibility. Correctly deploying EAP can be difficult, however, since there can be a great number of
questions to sort through to select the right protocol options. The key to EAP's flexibility is its status
as a framework. New methods can be designed as new requirements arise, which enable the
development of new methods for use in wireless LANs.
The topic of wireless LAN security protocols is a broad one, and this chapter draws the line at the
detailed mechanics of how the protocols work on a bit-by-bit basis. A great deal of complexity will be
described qualitatively, without the use of detailed packet diagrams.
Cryptographic Methods
Selecting an EAP method is usually driven by the type of back-end authentication system in use.
Early EAP methods focused mainly on providing a channel to communicate with the authentication
server. Newer methods designed for wireless networks enable communication with the authentication
server, but also meet three major goals that are unique to wireless LANs:
Strong cryptographic protection of user credentials
By definition, wireless LAN media should be regarded as open. Any data sent over a wireless
network must be protected if it is to remain secure. Most EAP methods designed for wireless
LANs use TLS to provide cryptographic protection of credentials.
Mutual authentication
Early wireless LAN protocol design considered authentication to be something that users
performed when required by APs. Falling AP prices have enabled attackers to deploy "rogue"
APs to steal user credentials. Common sense dictates in addition to user authentication, client
devices must be able to validate that they are connecting to a network that is what it appears
to be.
Key derivation
WEP with manually-defined keys provides very little protection of frames on the radio link.
Stronger security protocols need to use dynamic keys that are derived from an entropy pool.
One of the side effects of providing strong cryptographic protection is that such protocols also
generate a shared stream of cryptographically secure bits that can be used to distribute keys to
link-layer security protocols.
LEAP
The first widely-used method for authenticating wireless networks was Cisco's proprietary
Lightweight EAP (LEAP).[*] LEAP was a huge step forward from manual-key WEP, though it leaves a
great deal to be desired. In essence, LEAP is two MS-CHAP version 1 exchanges. One authenticates
the network to the user, and the second authenticates the user to the network. Dynamic keys are
derived from the MS-CHAP exchanges.
[*]
Some standards may refer to LEAP as "EAP-CiscoWireless."
Many of the worst security problems stem from the use of MS-CHAP version 1, which has numerous
security problems, and is subject to dictionary attacks. Code to exploit LEAP's weaknesses is now
widely available.
LEAP was an interim solution that had significant security benefits when compared to manually keyed
WEP, but the use of the antiquated MS-CHAP at its core limited its useful life. Once other protocols
became available, LEAP's purpose was extinguished. Cisco now recommends using PEAP or EAPFAST.
Code 13: EAP-TLS
Transport Layer Security (TLS) was designed for use on links subject to eavesdropping. TLS is the
standards-based successor to the Secure Socket Layer (SSL), the protocol that enabled secure web
transactions. In many respects, the use case for wireless LANs is similar to the Internet. Data must
be transmitted over a totally untrusted network with all manner of attackers. Establishing a trusted
communication channel over an untrusted network is the purpose of TLS.
TLS provides mutual authentication through certificate exchange. The user is required to submit a
digital certificate to the authentication server for validation, but the authentication server must also
supply a certificate. By validating the server certificate against a list of trusted certificate authorities,
the client can be assured that it is connecting to an network authorized by the certificate authority.
EAP-TLS was the first authentication method to meet all three goals for wireless networks.
Certificates provide strong authentication of both the users to the network, and the network to users.
Mutual authentication provides a strong guard against so-called "rogue" access points by enabling
clients to determine that an AP has been configured by the right department, rather than an attacker
who is intent only on stealing passwords. TLS also establishes a master secret that can be used to
derive keys for link layer security protocols.
Though secure, EAP-TLS has seen only limited use. Any potential user of the wireless network must
possess a digital certificate. Generating and distributing certificates while following processes to
verify trust is a major challenge. Organizations that already had a public key infrastructure were able
to use EAP-TLS easily; many organizations have shied away from building a PKI, and use an
alternative method.
Code 21: EAP-TTLS and Code 25: EAP-PEAP
Practically speaking, the requirement for PKI was a major bar to the use of strong authentication on
wireless LANs. PKIs are a significant undertaking in both technology and process. Most organizations
found it desirable to re-use existing authentication systems, such as a Windows domain or Active
Directory, LDAP directory, or Kerberos realm. Re-using existing accounts is much easier than creating
a parallel authentication system. The two EAP methods intended to enable the use of so-called
"legacy authentication methods" are Tunneled TLS (TTLS) and Protected EAP (PEAP).
Both TTLS and PEAP work in a similar fashion. In the first step of the protocol, they establish a TLS
tunnel using routines similar to EAP-TLS. Digital certificates on the authentication server are used to
validate that the network should be trusted before proceeding to the second step. In the second step,
the TLS tunnel is used to encrypt an older authentication protocol that authenticates the user to the
network. The first step is sometimes referred to as the "outer" authentication, since it is a tunnel that
protects the second or "inner" authentication.
Certificates are still required, but only for the outer authentication. Reducing the number of
certificates from hundreds or thousands down to a handful for authentication servers has made both
TTLS and PEAP much more popular than EAP-TLS. Within an organization, the small number of
certificates can be generated by a small certificate authority, rather than relying on expensive
certificates signed by an external certificate authority.
The slight difference between TTLS and PEAP is in the way the inner authentication is handled. TTLS
uses the encrypted channel to exchange attribute-value pairs (AVPs), while PEAP uses the encrypted
channel to start a second EAP exchange inside of the tunnel. The use of AVPs makes TTLS much
more flexible because AVPs can be used to run authentication methods that do not have
corresponding EAP methods.
One minor advantage to the use of TTLS and PEAP is that the inner and outer authentications can use
distinct usernames. Rather than revealing the username in unencrypted frames, both protocols can
submit an anonymous username for the outer authentication, and only reveal the user's true identity
through the encrypted channel. Not all client software supports the use of identity hiding.
Noncryptographic EAP Methods
Several EAP methods are not suitable for use directly on wireless networks without strong
cryptographic protection. However, they may be useful as inner authentication methods with PEAP or
TTLS.
Code 4: MD-5 Challenge
The MD-5 Challenge is used to implement the EAP analog of the CHAP protocol, specified in RFC
1994. Requests contain a challenge to the end user. For successful authentication, CHAP requires
that the challenge be successfully encoded with a shared secret. All EAP implementations must
support the MD-5 Challenge. It is not, however, widely supported on wireless networks because it
does not provide dynamic keys on wireless networks.
Code 6: Generic Token Card
Token cards such as RSA's SecurID and Secure Computing's Safeword are popular with many
institutions because they offer the security of "random" one-time passwords without the hassle of a
one-time password (OTP) rollout. The Request contains the Generic Token Card information
necessary for authentication. The Type-Data field of the request must be greater than zero bytes in
length. In the Response, the Type-Data field is used to carry the information copied from the token
card by the user. In both Request and Response packets, the Length field of the EAP packet is used
to compute the length of the Type-Data request.
EAP-GTC was standardized along with EAP in RFC 2284. It allows the exchange of cleartext
authentication credentials across the network. In addition to use with token cards, EAP-GTC is often
used in practice as an EAP method for "username+password" authentication. If the existing user
accounts are in a database that has one-way encrypted passwords (or compare-only passwords),
EAP-GTC provides an EAP method that can validate users. Naturally, if EAP-GTC is used to transport
reusable passwords, it must be used inside a tunnel for protection and server authentication.
Code 29: EAP-MSCHAP-V2
Microsoft CHAP version 2 (MS-CHAP-V2) was initially introduced with Windows 2000 and is
documented in RFC 2759. It was designed to address the shortcomings of MS-CHAP by eliminating
the weak encoding of passwords for older clients, providing mutual authentication, and improving
keying and key generation.
MS-CHAP-V2 is widely supported by Microsoft clients, and is commonly supported and used as an
inner authentication method with PEAP. MS-CHAP-V2 is the most common inner method used with
Windows domains. When used as an EAP method, EAP-MSCHAP-V2 can be used with either TTLS or
PEAP.
Code 18: EAP-SIM and Code 23: EAP-AKA
Two notable EAP methods working through the standards process are EAP-SIM and EAP-AKA, which
can be used for authentication against mobile telephone databases. EAP-SIM provides an interface to
the Subscriber Identity Module (SIM) database on GSM telephone networks. EAP-AKA is based on the
authentication system in third-generation mobile telephone networks, called Authentication and Key
Agreement (AKA).
EAP-SIM and EAP-AKA are useful for telecommunications companies that are interested in providing
integrated billing with mobile telephone accounts. Rather than require that users generate new
passwords to authenticate to data networks, they can use an existing smart chip and user account.
Other Inner Authentication Methods
TTLS is not restricted to using an EAP method for inner authentication. As a result, some older
methods may be used with TTLS. For some networks, the user database is stored in such a way that
there is no EAP method that provides a usable interface.
Password Authentication Protocol (PAP)
PAP was originally specified for use with PPP in RFC 1334. PAP transmits the username and password
across the network unencrypted. When used with wireless networks, PAP should not be used directly,
but only as an inner method inside of TTLS to ensure that the password is not revealed.
PAP can be used with any type of authentication system. Network logons can be validated against a
network operating systems. Token card servers can validate cleartext token codes. PAP is also useful
for one-way encrypted passwords that can be compared, but not read. One-way encrypted
passwords are used in Unix /etc/passwd files, as well as most LDAP directories. Kerberos systems
also store passwords with one-way encryption.
Challenge Handshake Authentication Protocol (CHAP)
Like PAP, CHAP was originally designed for use with PPP, and is specified in RFC 1994. In CHAP, the
authentication server challenges the client, and the client proves that it is in possession of the shared
secret by successfully responding to the challenge.
CHAP was originally designed to avoid sending secrets, such as passwords, in the clear. The
downside of CHAP is that it requires the password in cleartext at both ends of the link. On the server
end, the password must either be stored in the clear, or reversibly encrypted to enable server
software to recover the cleartext.
CHAP is not typically used in wireless environments unless a large user database already exists and is
using CHAP.
MS-CHAP, version 1
MS-CHAP was designed by Microsoft to offer similar functionality to CHAP, but with enhanced
functionality for Windows systems. It is proprietary to Microsoft, but documented in RFC 2433. Unlike
CHAP, MS-CHAP does not require that the shared secret be stored in cleartext at both ends of the
link. Instead of using the cleartext password as the shared secret, MS-CHAP uses the MD4 hash of
the user password.
MS-CHAP is useful in environments where Microsoft authentication databases are used. However,
there are particular issues with MS-CHAP that make it undesirable from a security point of view; it
should only be used by network managers who must support very old Microsoft clients, such as
Windows 95/98. MS-CHAP is not an EAP method, and is only supported by TTLS.
EAP Method Filtering
Some EAP methods are stronger than others. In the interests of "security," some
equipment vendors filter EAP methods in the authenticator. EAP-MD5 is often selected as
a target for EAP type filtering because it is much weaker than TLS-based methods.
EAP method filtering is not necessary, and has the potential to get in the way of good
security practice. Many filters only allow TLS-based tunnel methods (types 13, 21, and
25). If a new EAP method is defined, many EAP method filter implementations will block
it by default. Allowing configurable EAP method filters would be one potential solution,
but I do not know of any such implementations.
In practice, EAP method filters are not that valuable from a security perspective.
Authentication servers can enforce the use of strong authentication by refusing any
requests using weak methods. In any case, the lack of keying material may prevent
network connection. If no keys can be derived from the authentication, many APs will
void the authentication, kick the client off, and start over. EAP-MD5 authentications will
start, succeed, fail to key, and be restarted when the AP dissassociates the client.
Due in part to these problems, EAP method filtering is not allowed by the EAP
specification. Section 2.3 of RFC 3748 mandates that authenticators must forward any
EAP methods to the authentication server for processing.
802.1X: Network Port Authentication
As LAN acceptance mushroomed in the 1990s, LAN ports popped up everywhere. Some types of
organizations, such as universities, were further hampered by a need for openness. Network
resources must be made available to a user community, but that community is fluid. Students are
not like many network users. They frequently move from computer to computer and do not have a
fixed network address; they may also graduate, transfer, enroll, leave campus, work on staff, or
undergo any number of changes that may require changes in access privileges. Although network
access must be extended to this fluid community, academic budgets are frequently tight, so it is
important to prevent unauthorized use by outsiders.
In short, a generic network sign-on was required. Academic environments would not be the sole
beneficiaries, however. Authentication to access network resources is common among Internet
service providers, and corporations found the idea attractive because of the increasing flexibility of
staffing plans.
Authentication to network devices at the link layer is not new. Network port authentication has been
required by dial-up access servers for years. Most institutions already have a wide range of deployed
infrastructure to support user authentication, such as RADIUS servers and LDAP directories. PPP over
Ethernet (PPPoE) could conceivably be used to require user authentication to access an Ethernet, but
it would add an unacceptable level of encapsulation overhead and complexity. Instead, the IEEE took
the PPP authentication protocols and developed LAN-based versions. The resulting standard was
802.1X, "Port-Based Network Access Control."
802.1X Architecture and Nomenclature
802.1X defines three components to the authentication conversation, which are all shown in Figure
6-6 (a). The supplicant is the end user machine that seeks access to network resources. Network
access is controlled by the authenticator; it serves the same role as the access server in a traditional
dial-up network. Both the supplicant and the authenticator are referred to as Port Authentication
Entities (PAEs) in the specification. The authenticator terminates only the link-layer authentication
exchange. It does not maintain any user information. Any incoming requests are passed to an
authentication server, such as a RADIUS server, for actual processing.
Ports on an 802.1X-capable device are in an authorized state, in which the port is enabled, or an
unauthorized state, in which it is disabled. Even while in the unauthorized state, however, the
specification allows DHCP and other initialization traffic if permitted by a network manager.
The authentication exchange is logically carried out between the supplicant and the authentication
server, with the authenticator acting only as a bridge. Figure 6-6 (b) shows the logical protocol
architecture. From the supplicant to the authenticator (the "front end"), the protocol is EAP over
LANs (EAPOL), as defined by 802.1X. On the "back end," EAP is carried in RADIUS packets. Some
documentation may refer to it as "EAP over RADIUS." The supplicant is carrying on an EAP exchange
with a RADIUS server, even though the port is unauthorized and the supplicant is not using an IP
address.
Figure 6-6 can be read as two different scenarios. In the enterprise scenario, the supplicant is a
corporate host on the edge of the enterprise network, and the RADIUS server is located in the
enterprise core. The figure also depicts an ISP using 802.1X to authenticate users, in which case the
lefthand side of the figure is an ISP access area, and the righthand side is the ISP backbone.
Figure 6-6. 802.1X architecture
One of the advantages of using RADIUS is that it has a great deal of support for many different user
databases. In addition to local databases, a RADIUS server can be used as a gateway to LDAP
directories, Unix authentication such as NIS or PAM, Kerberos realms, Microsoft Windows user
accounts, or even other RADIUS servers. RADIUS is quite flexible, and can be designed to
accommodate a wide range of user databases, and can even serve to integrate several disparate user
databases into a unified form. There are some restrictions when using RADIUS to provide
authentication based on Windows domains or Active Directory that will be discussed in Chapter 22.
802.1X is a framework, not a complete specification in and of itself. The actual authentication
mechanism is implemented by the authentication server. 802.1X supplies a mechanism for issuing
challenges and confirming or denying access, but it does not pass judgment on the offered
credentials. Changes to the authentication method do not require complex changes to the end user
devices or the network infrastructure. The authentication server can be reconfigured to "plug in" a
new authentication service without changes to the end user driver software or switch firmware.
802.1X frame filtering
802.1X prevents network access by unauthorized users by requiring authentication before allowing
traffic to pass. Ports in the unauthorized state are usually restricted to sending authentication frames
only, while traffic is dropped. The 802.1X standard has a formal discussion of how to accomplish this
process. For the purposes of this book, however, it suffices to say that unauthorized ports will drop
any non-EAPOL frames. Once the station has successfully authenticated, data frames are sent to the
appropriate network.
EAPOL Encapsulation
The basic format of an EAPOL frame is shown in Figure 6-7. EAPOL encapsulation is now analyzed by
many popular network analyzers, including Ethereal. The frame's components are:
MAC header
Figure 6-7 shows the encapsulation on both wired Ethernet and 802.11. The two MAC headers
differ, although the payload of the 802.1X frame is identical.
Figure 6-7. EAPOL frame format
Ethernet Type
As with any other Ethernet frame, the Ethernet Type field contains the two-byte type code
assigned to EAPOL: 88-8e.
Version
Version 1 was standardized in the 2001 version of 802.1X; version 2 was specified in 802.1X2004. This chapter describes the 2001 version, which is much more commonly implemented.
Packet Type
EAPOL is an extension of EAP. In addition to the EAP messages that are described in the
previous section, EAPOL adds some messages to adapt EAP to the port-based LAN
environment. Table 6-2 lists the packet types and their descriptions.
Table 6-2. EAPOL message types
Packet
type
Name
Description
0000
0000
EAP-Packet
Contains an encapsulated EAP frame. Most frames are EAP-Packet
frames.
0000
0001
EAPOL-Start
Instead of waiting for a challenge from the authenticator, the
supplicant can issue an EAPOL-Start frame. In response, the
authenticator sends an EAP-Request/Identity frame.
0000
0010
EAPOL-Logoff
When a system is done using the network, it can issue an EAPOLLogoff frame to return the port to an unauthorized state.
0000
0011
EAPOL-Key
0000
0100
EAPOLEncapsulated-ASFAlert
EAPOL can be used to exchange cryptographic keying information.
The Alerting Standards Forum (ASF) has defined a way of allowing
alerts, such as SNMP traps, to be sent to an unauthorized port using
this frame type.
Packet Body Length
This two-byte field is the length of the Packet Body field in bytes. It is set to 0 when no packet
body is present.
Packet Body
This variable-length field is present in all EAPOL frames except the EAPOL-Start and EAPOLLogoff messages. It encapsulates one EAP packet in EAP-Packet frames, one key descriptor in
EAPOL-Key frames, and one alert in EAPOL-Encapsulated-ASF-Alert frames.
Addressing
In shared-media LANs such as Ethernet, supplicants send EAPOL messages to the group address of
01:80:C2:00:00:03. On 802.11 networks, ports do not exist as such, and EAPOL can proceed only
after the association process has allowed both the supplicant (mobile wireless station) and the
authenticator (access point) to exchange MAC addresses. In environments such as 802.11, EAPOL
requests use station addresses.
802.1X on Wireless LANs
802.1X provides a framework for user authentication over any LANs, including wireless. For the
purposes of this book, the "port" in 802.1X on wireless is an association between a wireless device
and its access point. The successful exchange of Association Request and Association Response
frames is reported to the 802.1X state engine as the link layer becoming active. Once associated, a
station can exchange 802.1X frames in an attempt to become authorized. The completion of the
802.1X authentication exchange, including key distribution, is reported to the user as the interface
coming up.
Sample 802.1X Exchange on 802.11
EAPOL exchanges look almost exactly like EAP exchanges. The main difference is that supplicants can
issue EAPOL-Start frames to trigger the EAP exchange, and they can use EAPOL-Logoff messages to
deauthorize the port when the station is done using the network. The examples in this section
assume that a RADIUS server is used as the back-end authentication server, and therefore they show
the authenticator performing translation from EAP on the front end to RADIUS on the back end. EAP
authentication in RADIUS packets is specified in RFC 2869.
This example exchange also shows the use of EAPOL-Key frames to distribute key information for link
layer security protocols. Figure 6-8 shows a sample EAPOL exchange on an 802.11 network. The
figure shows a successful authentication, whose steps are:
1. The supplicant associates with the 802.11 network. Association is a simple two-frame exchange
which nearly always succeeds.
2. The supplicant starts the 802.1X exchange with an EAPOL-Start message. This step is optional.
Not all supplicants send EAPOL-Start messages, so this step may not be present.
3. The "normal" EAP exchange begins. The authenticator (access point) issues an EAPRequest/Identity frame. Request/Identity frames may be sent without first having an EAPOLStart if the access point only forwards frames for authenticated sessions. Unsolicited
Request/Identity frames indicate to the supplicant that 802.1X authentication is required.
4. The supplicant replies with an EAP-Response/Identity frame, which is passed on to the RADIUS
server as a Radius-Access-Request packet.
5. The RADIUS server determines the type of authentication that is required, and sends an EAPRequest for the method type. The EAP-Request is encapsulated in a Radius-Access-Challenge
packet to the AP. When it reaches the AP, the EAP-Request is passed on to the supplicant. EAP
Requests are are often denoted EAP-Request/Method, where the Method refers to the EAP
method in use. If PEAP is in use, the return packet will be written as EAP-Request/PEAP.
6. The supplicant gathers the reply from the user and sends an EAP-Response in return. The
response is translated by the authenticator into a Radius-Access-Request with the response to
the challenge as a data field.
Steps five and six repeat as many times as is necessary to complete the authentication. If it is
7.
an EAP method that requires certificate exchange, multiple steps are almost certainly required.
Many EAP exchanges can require 10-20 round trips between the client and RADIUS server.
7. The RADIUS server grants access with a Radius-Access-Accept packet, so the authenticator
issues an EAP-Success frame and authorizes the port. Authorization may depend on parameters
passed back from the RADIUS server.
8. Immediately following receipt of the Access-Accept packet, the access point distributes keys to
the supplicant using EAPOL-Key messages. Key distribution is discussed in the next chapter.
9. Once keys are installed in the supplicant, it can begin sending data frames to access the
network. It is quite common at this point for DHCP configuration to take place.
10. When the supplicant is done accessing the network, it sends an EAPOL-Logoff message to put
the port back into an unauthorized state.
Figure 6-8. Typical 802.1X exchange on 802.11
Exchanges similar to Figure 6-8 may be used at any point. It is not necessary for the user to begin
an EAPOL exchange with the EAPOL-Start message. At any point, the authenticator can begin an
EAPOL exchange by issuing an EAP-Request/Identity frame to refresh the authentication data. Re-
authentications are often triggered by session timeout values to refresh keys.
Dynamic keying
The EAPOL-Key frame allows keys to be sent from the access point to the client and vice versa. Key
exchange frames are sent only if the authentication succeeds; this prevents the compromise of key
information. EAPOL-Key frames can be used periodically to update keys dynamically as well. Several
of the weaknesses in WEP stem from the long lifetime of the keys. When it is difficult to rekey every
station on the network, keys tend to be used for long periods of time. Several experts have
recommended changing WEP keys on a regular basis, but no practical mechanism to do so existed
until the development of 802.1X.
Chapter 7. 802.11i: Robust Security
Networks, TKIP, and CCMP
802.1X provides a framework for authentication and key management, which addresses two of the
major flaws in the design of WEP. The major remaining flaw to be addressed is the lack of
confidentiality provided by WEP encryption. Fixing link layer encryption was taken on by Task Group I
of the 802.11 working group. In June 2004, their work was finally complete when the standard was
ratified, after several delays.
802.11i takes a two-track approach to addressing the weaknesses in link-layer encryption. Its major
components are two new link-layer encryption protocols. The first, the Temporal Key Integrity
Protocol (TKIP) was designed to bolster security to the greatest extent possible on pre-802.11i
hardware. The second, Counter Mode with CBC-MAC Protocol (CCMP), is a new encryption protocol
designed from the ground up to offer the highest level of security possible.
The Temporal Key Integrity Protocol (TKIP)
The first new link layer encryption protocol to be widely implemented was the Temporal Key Integrity
Protocol (TKIP).[*] The major motivation for the development of TKIP was to upgrade the security of
WEP-based hardware. Typically, chipsets capable of WEP offered hardware support for RC4
encryption. With the heavy lifting of encryption implemented in hardware, software and firmware
upgrades make the rest possible. TKIP retains the basic architecture and operations of WEP because
it was designed to be a software upgrade to WEP-based solutions.
[*]
TKIP was initially called "WEP2" as it worked through the standards organizations. When WEP was shown to be fundamentally
flawed, the protocol was renamed to try to distinguish it from WEP.
TKIP Differences from WEP
TKIP incorporates several new protocol features to defend WEP's weak points against attack. TKIP
retains WEP's basic architecture and operation, but adds "safety belts" around WEP's most vulnerable
points:
Key hierarchy and automatic key management
Rather than take the WEP approach of only a single master key that is used directly, TKIP uses
master keys. The keys that are ultimately used to encrypt frames are derived from master
keys. TKIP was also developed with key management operations so that master keys can be
refreshed in a secure manner.
Per-frame keying
Although TKIP retains the RC4-based frame encryption of WEP, it derives a unique RC4 key for
each frame (from the master key) to mitigate attacks against weak WEP keys. The process by
which a unique key is derived for each key is called key mixing.
Sequence counter
By numbering each frame with a sequence number, out-of-order frames can be flagged,
mitigating against reply attacks, in which attackers capture valid traffic and re-transmit it at a
later time.
New message integrity check (MIC)
TKIP replaces WEP's linear hash with a more robust cryptographic integrity check hashing
algorithm called Michael. More robust hashing makes it easier to detect frame forgeries.
Additionally, the source address is among the items protected by the integrity check, which
makes it possible to detect forged frames that claim to be from a particular source.
Countermeasures on message integrity check failures
TKIP was designed to be implemented on existing hardware, and suffers from a number of
limitations. Michael can be compromised in an active attack with relative ease, so TKIP includes
countermeasures to limit the damage from an active attack.
TKIP is also commonly used in tandem with key management protocols based on 802.1X, which
enables the TKIP master keys to be derived from authentication transactions.
TKIP initialization vector use and key mixing
WEP's construction of WEP seeds from the initialization vector (IV) and the WEP key introduces major
weaknesses. By simply concatenating the IV and the key to build the seed, the IV itself reveals a
significant amount of key structure. Attackers can notice that IV re-use, and note that therefore, the
key stream used to encrypt the frame is identical. With a 24-bit IV, the IV space lasts for 16 million
frames. On a busy network, 16 million frames is not very many. To make matters worse, the IV
space is specific to the scope of the key in use. On a network that uses static WEP, the IV space is
shared by all stations on the network. Finally, and most famously, the IV is the initial 24 bits of the
key, and makes WEP keys vulnerable to recovery with the Fluher/Mantin/Shamir attack discussed in
Chapter 5.
To mitigate the attacks against initialization vectors, TKIP doubles the length of the IV from 24 to 48
bits. This increases the size of the initialization vector space from 16 million to 281 trillion, which
effectively prevents exhausting the IV space during the limited lifetime of a key.
TKIP also performs key mixing to counter attacks against WEP. Key mixing changes the RC4 key
used to encrypt each frame. Every frame in TKIP is encrypted with an RC4 key unique to that frame.
Key mixing, which will be discussed in more detail later, further extends the initialization vector
space. By incorporating the sender's MAC address into the mixing calculation, two stations can use
the same initialization vector, yet still derive different RC4 keys to encrypt frames. Key mixing also
helps to counter the Fluhrer/Mantin/Shamir attack. Successful application of the principles of the
attack require a collection of "weak keys" with the same secret bits. By changing the key for every
frame, TKIP prevents an attacker from gathering enough data to attack any of the per-frame keys.
TKIP sequence counter and replay protection
In addition to its larger size, the TKIP initialization vector serves as a sequence counter. When a new
master key is installed, the initialization vector/sequence counter is set to one. Each frame
transmitted increments the sequence counter by one.
To defend against replay attacks, TKIP maintains the most recent sequence counter value received
from each station. When a frame is successfully received, the sequence counter is checked against
the most recently received sequence counter. If it is larger than any previous value, the frame is
accepted. If it is smaller than the most recently received sequence counter value, it is rejected.
Task group E is developing quality of service extensions. Current 802.11e drafts contain a block
acknowledgment protocol feature that uses a single frame to acknowledge multiple transmissions. In
the January 2005 draft of 802.11e, re-ordering due to block acknowledgments must complete before
applying replay detection.
802.11 unicast frames must be acknowledged. If either the original frame or its acknowledgment is
lost, the frame will be retransmitted. At the receiver, this may result in the same sequence counter
being received twice. Duplicate sequence numbers may simply represent evidence of an error on the
link and do not necessarily indicate an active attack in progress.
The Michael integrity check and countermeasures
WEP's integrity check is a linear hash value, which is totally unsuitable for cryptographic applications.
One of the major design challenges faced by TKIP was strengthening the integrity check while
retaining reasonable performance. Most 802.11 chipsets in use at the time TKIP was designed used a
relatively low-power processor, and were thus not capable of performing mathematical operations
fast enough to do integrity checks at "air-speed." Michael is implemented entirely with bitwise
operations such as swaps, shifts, or even discarding bits. As a result, it can run without deleterious
performance effects, even on the tiny microprocessors found in most 802.11 interfaces. (Michael is
discussed in detail later in this chapter.)
The corollary to Michael's design is that it does not provide a great deal of security. It is significantly
better than a cyclic redundancy check, but it does not offer security against a sustained and
determined attack. TKIP incorporates countermeasures to detect and respond to an active attack by
shutting down the network and refreshing the keys in use.
Michael's countermeasures are an admission of the weakness of the underlying cryptographic
fundamentals. Design of Michael was constrained by the need to be backwards-compatible with most
of the existing RC4-based hardware on the market. Michael's design is elegant and accomplishes a
great deal within its severe design constraints, but it is still built on a foundation of sand. Time will
tell whether Michael is sufficiently secure; I would not be surprised to see a major attack against it in
the next year. However, the improved protocol design of TKIP limits the negative consequences of an
attack. I expect that a compromise of Michael would result in denial-of-service attacks, but would not
lead to wholesale network compromise.
TKIP Data Processing and Operation
Like WEP, TKIP provides support for encryption and integrity protection as part of the same process,
as shown in Figure 7-1. TKIP's design as a set of safety features around WEP is quite clear from the
diagram.
As input, TKIP takes the following items:
The frame, naturally.
A temporal key used to encrypt the frame.
A MIC key used with Michael to protect the frame contents. TKIP derives a pair of keys so that
the station-to-AP MIC key is different from the AP-to-station MIC key. One of the ways that
TKIP differs from WEP is that the MIC uses a key.
The transmitter address is used as an input to TKIP because it is required to perform origin
authentication. The transmitter address is supplied with the frame and does not need to be
supplied by higher level software.
A sequence counter, maintained by the driver or the firmware.
Figure 7-1. TKIP frame processingencryption
TKIP key mixing and key construction
TKIP derives a unique key for each frame transmitted. The key is derived from the initialization
vector/sequence counter, the address of the transmitter of the frame (which may not be the source
of the frame), and the temporal key. Key mixing ensures that the key used varies significantly from
one frame to the next, and prevents any attacks which assume that the secret component of the WEP
key are constant from frame to frame. By including the transmitter address in the mixing
computation, two stations can use the same initialization vector yet derive different two RC4 keys.
Limited processing power of the 802.11 controllers dictated the design of the key mixing function,
which is shown in Figure 7-2. TKIP splits the computation of the mixed key into two phases. Phase
one takes, as input, the transmitter address, the high-order 32 bits of the sequence counter, and the
128-bit temporal key. As output, it gives an 80-bit value. Although the computation is somewhat
involved, it consists entirely of "easy" operations such as addition, shifts, and exclusive ORs, to
reduce the computational load. The phase one value is constant as long as the 32 high-order bits of
the sequence counter are constant, so it only needs to be computed once every 65,536 frames.
Phase two of the key mixing function must be computed for every frame. As input, phase two takes
the phase one result, the temporal key, and the 16 low-order bits of the sequence counter. The only
input that changes from frame to frame is the sequence counter. It changes in a well-defined way, so
implementations can and do pre-compute values based on the next set of sequence counter values
that will be needed as frames are queued for transmission.
Figure 7-2. TKIP key mixing
The output of the phase two of the key mixing process is a 128-bit RC4 key that can be used as a
WEP seed. The 16 low-order bytes are used to generate a WEP IV. The middle byte of the WEP IV is
constructed so that it avoids generating weak RC4 keys. Many 802.11 interfaces incorporate
hardware assistance that can take an RC4 key as input and generate and apply the resulting
keystream to encrypt the frame. The output of the phase two mixing process can be given directly to
an 802.11 interface that supports this type of hardware assistance.
TKI P data transmission
When a frame is generated and sent to TKIP for transmission, the following sequence of events
occur.
1. The 802.11 frame is queued for transmission. It consists of a frame header and the payload.
Like WEP, TKIP protects only the payload of the 802.11 MAC, and leaves the 802.11 frame
header, as well as any lower-layer headers, intact.
2. The Message Integrity Check (MIC) is computed. Unlike WEP, TKIP's MIC is a more robust
cryptographic hash. It uses a secret key as part of its validation process, and protects much
more than just the 802.11 frame payload. In addition to the frame data, the MIC incorporates
the source and destination address as well as the addition of priority bits that will be used by
the future 802.11e standard.
3. Sequence numbers are assigned to fragments. Unlike WEP initialization vectors, TKIP's
sequence counter increases by one with each fragment. If no fragmentation is necessary, only
one sequence counter increment will be necessary. In the case of multiple fragments, the
4.
3.
counter will increment multiple times.
4. Each frame is encrypted with a unique per-frame WEP key. By carrying out the key mixing
functions, TKIP derives the WEP key used for the frame. The per-frame key is passed to WEP as
an IV plus a secret key; both components will change for every frame.
5. The frame plus Michael message integrity check value from step 2 and the RC4 key from step 4
are passed to WEP, which encapsulates the frame as described in Chapter 5. Note that this
means that a TKIP-protected frame will include WEP components as well.
TKIP frames have WEP-like encapsulation. Figure 7-3 shows the encapsulation of a frame that has
been protected by TKIP. The IV/KeyID is retained from WEP, but has a different interpretation. The
first three bytes carry part of the TKIP sequence counter, and identify the key number in use.
Although TKIP can support multiple keys, in practice, only KeyID 0 is distributed. An Extended IV
(EIV) field is used to carry the remainder of the sequence counter. Following the data payload, TKIP
appends the MIC. WEP adds its own integrity check value, to retain WEP in as unchanged a form as
possible.
Figure 7-3. TKIP encapsulation
TKIP reception
When TKIP receives a frame, the encryption and transmission process must be reversed. TKIP's
design as an expanded version of WEP adds several new checkpoints in the decryption process, and
therefore it is worth treating in more detail. Figure 7-4 shows a block diagram, with all of the ways
that frames may be discarded before passing up to higher protocol layers.
1. When a frame is received by the wireless interface and passes the frame check sequence to
ensure it has not been corrupted, it is passed to TKIP for validation.
2. The first step TKIP takes is to check the sequence number to prevent replay attacks. TKIP's
replay protection requires that frames be received in strict order. Frames that have a sequence
number equal to or lower than the last received valid frame are discarded as a replay attempt.
This is stricter than some other technologies, which maintain a replay window and allow
"recent" arrivals to be somewhat out of order. For a link layer technology, however, it makes
sense to insist on strict ordering because there is only a small possibility of out-of-order delivery
on a single hop.
3. The WEP seed used to encrypt the packet is recovered. With the transmitter address, temporal
key, and sequence counter, the receiver can unmix the key to recover the WEP seed.
4. With the WEP seed in hand, the outer WEP layer around the frame can be removed and the
3.
4.
contents recovered. As part of this process, the WEP ICV is checked. Although it is a weak
integrity check, it can be used to protect against applying countermeasures unnecessarily.
5. If fragmentation was applied, it may be necessary to wait for further frames to arrive before
reassembling a complete payload. Fragmentation is not used often on 802.11, however.
6. Once the frame is reassembled, Michael is calculated over the contents of the frame. If the
calculated MIC matches the MIC on the packet, the frame is passed to higher layers, and the
sequence counter is set to the value of the sequence counter on the frame. If the MIC check
fails, countermeasures are triggered.
Figure 7-4. TKIP reception process
One of the reasons for retaining the WEP encapsulation is that it enables greater backwardscompatibility with existing hardware. It also prevents unnecessary triggers of MIC failures. Frames
that are corrupted due to noise on the radio link will likely fail either the 802.11 frame check
sequence or the WEP integrity check value, and will be discarded before triggering a MIC failure.
The Michael Integrity Check
One of WEP's weakest points was the integrity check, which was supposed to ensure that the frame
was not tampered with as it traversed the wireless medium. WEP adopted the cyclic redundancy
check (CRC), which proved woefully inadequate. One of the major goals of TKIP was to design a
message integrity check (MIC) with much more solid cryptographic underpinnings. The resulting
algorithm, called Michael, is a compromise in many ways. Michael is stronger than a simple linear
hash, but the desire of the standards committee to make it easy to implement with low overhead
placed severe constraints on its design.
From an architectural point of view, Michael plugs in to the MAC service layer. That is, Michael
operates on frames passed down to it from higher-layer protocols. Michael does not protect individual
802.11 frames; it protects the reassembled data unit given to 802.11 for transmission. [*] Part of the
reason for implementing Michael above the MAC layer as opposed to within it is that it leaves
products the flexibility to implement Michael in dedicated hardware or in driver software running on
the device. Older chipsets and 802.11 interfaces did not build in specialized hardware assistance for
Michael because it had not been developed; these products can implement Michael with updated
drivers or host software. Newer devices designed with emerging standards in mind are able to build
Michael support into the chipset if desired, and will not need driver support that is as extensive.
[*]
In practice, the distinction between the MAC service layer and the MAC protocol layer is not particularly necessary. Although
802.11 can accommodate frames over 2,000 bytes, most commercial implementations limit frame size to be compatible with
Ethernet to avoid higher-layer fragmentation.
Several attacks discussed previously served as the motivation for Michael. Two of the most notable
are the bit flipping attack and a whole class of header modification. The former attack depends on
the weak cryptographic properties of the CRC. As a linear hash algorithm, it is well known how
changing the CRC input bits will change the output. Attackers could potentially flip bits in the frame
to change its content, and change a bit in the WEP integrity check to compensate. In header
modification attacks, malicious attackers may forge the source or transmitter address, or attempt to
redirect frames by modifying the destination address.
Make no mistake, Michael is not a particularly secure cryptographic protocol. It was designed to offer
breathing room to users who have an extensive installed base while they upgrade the security of
their networks. It is only a short-term fix while the long-term solution is developed.
Michael data processing
Michael operates on higher-level frames passed down to the MAC. When a higher-level frame is
queued for transmission, one of the first transmission steps is to compute the message integrity
check (MIC) value. Michael operates on the input shown in Figure 7-5 (a). In addition to the
destination address (DA) and source address (SA), Michael adds four zero bytes before the
unencrypted data. The last three bytes are reserved; the first is a priority field that is reserved for
future standardization work.
Michael operates on 32-bit blocks of data. Data is padded out with zeros so that it will be a multiple
of 32-bit blocks. Padding is used only for the computation of the MIC. It is not transmitted. Michael
computes the MIC value by working on successive 32-bit blocks of data.
Figure 7-5. Michael input
When complete, the MIC is added on to the tail of the data frame, and the data-plus-MIC is given to
802.11 for transmission. If it needs to be fragmented, it will be. During the fragmentation process,
the MIC value may be split across multiple 802.11 fragments, which is acceptable. In Figure 7-5, the
data-plus-MIC unit from higher layers requires fragmentation and is split across the frame boundary.
This is a perfectly acceptable occurrence because the two fragments will be reassembled before
validating the MIC. Figure 7-5 (b) illustrates this process. The input from Figure 7-5 (a) has been
split into two frames for transmission. Each layer will get its own MAC and PLCP headers, and will be
processed by the WEP layer in TKIP.
Michael countermeasures
Michael is not very resistant to attack, although it appears to be as good as one could hope, given
the constraints of a great deal of legacy hardware on the market. Michael is not able to withstand a
determined active attack, so it incorporates countermeasures against such attacks.
Countermeasures happen at the MIC check. To make it to that point, a frame must have passed
through the replay protection and the WEP integrity check value validation. Forging a frame to pass
the replay protection and the WEP integrity check is not a trivial matter, especially since the latter
value depends on the value of the mixed per-frame key.
If an attacker were able to bypass replay protection and the WEP integrity check, it would be possible
to mount a brute-force attack on the Michael integrity check. To guard against brute-force attempts
to pass Michael, Michael's countermeasures shut down communications to limit the number of active
attacks that can be made against any particular key. If an attack is sustained, countermeasures call
for refreshing the key and shutting down communications.
When a station detects a MIC failure, it executes the following procedure:
1. The MIC failure is noted and logged. Before the MIC is validated, the frame must pass through
the replay protection hurdle as well as the legacy WEP integrity check. Getting a frame to
Michael for validation is not a trivial undertaking. Therefore, any MIC validation error is likely to
be an extremely security-relevant matter that should be investigated by system administrators.
2. If the failure is the second one within a 60-second window, countermeasures dictate shutting
down communications for a further 60 seconds. When the second MIC failure within 60 seconds
is detected, all TKIP communication is disabled for 60 seconds. Instituting a communication
3.
2.
blackout makes it impossible for an attacker to mount a sustained attack quickly.
3. Keys are refreshed. Stations delete their copies of the master keys and request new keys from
the authenticator; authenticators are responsible for generating and distributing new keys.
Although 802.11i sets the Michael countermeasures time at 60 seconds, some vendors allow
configuration of the Michael countermeasures window to other values.
Counter Mode with CBC-MAC (CCMP)
TKIP is better than WEP, but that is the only statement that can be made with certainty. WEP's basis
in a stream cipher will always leave lingering doubts about the security of anything built using a
similar set of operations. To address the concerns of the 802.11 user community, the IEEE working
group began developing a security protocol based on the Advanced Encryption Standard (AES) block
cipher. AES is a flexible
What Is WPA?
Wi-Fi Protected Access (WPA) is a marketing standard put together by the Wi-Fi Alliance.
The Wi-Fi Alliance leaves the details of hammering out standards to other bodies like the
IEEE. As a trade association, however, they react to ensure that the perception of the
industry remains positive.
When the first cracks appeared in the foundation of WEP, the IEEE launched a working
group to develop improved security standards. Building secure cryptographic protocols is
difficult work, however, and 802.11i has been delayed repeatedly past its expected due
date. 802.11i specifies two new security protocols: TKIP and CCMP. TKIP was designed
to be backwards compatible with existing hardware at the time it was developed,
whereas CCMP was designed essentially from the ground up. As a result, TKIP was
finished well before CCMP was ready.
To address security concerns in the market, the Wi-Fi Alliance worked to speed up
deployment of TKIP by coming up with an interim marketing standard called WPA. WPA
version 1 is based on the third draft of 802.11i (from mid-2003); WPA version 2 is the
final standardized version of 802.11i from mid-2004.
WPA includes both authentication through 802.1X and encryption. It comes in two
flavors: WPA Personal, which is equivalent to pre-shared key authentication in 802.11i,
and WPA Enterprise, which uses the authenticated key mode that derives keys from TLS
entropy.
cipher that can operate at many key lengths and block sizes; to prevent user confusion, 802.11i
mandates the use of AES with both 128-bit keys and 128-bit blocks.
There has been some concern over the key size used with 802.11i. AES can operate with a variety of
key lengths. The U.S. National Security Agency has approved AES for use with "secret" data with
128-bit and longer keys, but more sensitive "top secret" data requires the use of 192-or 256-bit
keys.[*] Some observers have therefore concluded that 128-bit keys do not offer adequate security.
Regardless of the merits in the debate over key size, 128-bit AES is much better suited to 802.11
frame encryption RC4 at any key length.
[*]
See the Committee on National Security Systems (CNSS) Policy No. 15, Fact Sheet No. 1, "National Policy on the Use of the
Advanced Encryption Standard (AES) to Protection National Security Systems and National Security Information" at
http://www.nstissc.gov/Assets/pdf/fact%20sheet.pdf.
The link-layer security protocol based on AES is called the Counter Mode with CBC-MAC Protocol
(CCMP).[ ] The name comes from the underlying use of the block cipher, in the Counter Mode with
CBC-MAC (CCM) mode, which is specified in RFC 3610. CCM is a "combined" mode of operation, in
which the same key is used in encryption for confidentiality as well as creating a cryptographically
secure integrity check value.
[
] One of the delays in finishing 802.11i was that the AES algorithm was initially based on a different mode of operation, AESOCB. Intellectual property concerns required that it be dropped from the final specification in favor of CCMP.
In September 2003, the National Institute of Standards and Technology (NIST) began a study of
CCM. In May 2004, NIST gave its approval to CCM, which will allow 802.11i to serve as the basis for
secure wireless LANs in demanding applications.
CCMP Data Processing
Like other link-layer encryption methods, CCMP provides support for encryption and integrity
protection as part of the same process, as shown in Figure 7-6.
Figure 7-6. CCMP frame processingencryption
As input, CCMP takes the following items:
The frame, naturally.
A temporal key used to encrypt and authenticate the frame. One key is used for both frame
encryption and frame authentication.
A key identifier to support the use of multiple keys over the air. Only one key, however, will be
used for any given frame.
A packet number, used to uniquely identify the frame being transmitted. The packet number is
incremented for each frame transmission, but it remains the same for each attempt to transmit
the frame. That is, retransmissions do not cause the PN to increment.
CCMP data transmission
When a frame is generated and sent to TKIP for transmission, this procedure occurs:
1. The 802.11 frame is queued for transmission. It consists of a frame header and the payload.
Like WEP, TKIP protects only the payload of the 802.11 MAC, and leaves the 802.11 frame
header, as well as any lower-layer headers, intact.
2. A 48-bit Packet Number (PN) is assigned. As with the TKIP sequence number, packet numbers
are never re-used for the same temporal key. Packet numbers are increased by one for every
transmission, and are used for replay detection.
3. The Additional Authentication Data (AAD) field is constructed. It consists of the fields in the
frame header which must be authenticated for security but must also remain unencrypted so
that the 802.11 protocols may operate. The receiver will use the AAD field to verify that the
authenticated fields were not altered in transit. The AAD field protects the 802.11 protocol
version, frame type, the distribution system bits, and fragmentation and order bits. It also
protects the address fields from the MAC header, and the sequence control field, with the
sequence number (though not the fragment number) set to zero. It ends with two optional
components: the fourth address field from the MAC header (if the wireless distribution system is
in use), and quality of service header information.
4. Next, the CCMP nonce is built. Nonces are little bits of data that guarantee that the encryption
operation occurs on unique data. Nonces should never be re-used with the same key. In CCMP,
the nonce is constructed with the packet number and the sender's address so that the same
packet number can be used by multiple stations. The nonce also includes priority data used in
QoS.
5. Next, the CCMP header is built. It consists of the packet number, broken across six bytes, with
a key identifier in the middle. As in WEP, there are four potential key slots used in CCMP. The
Extended IV bit is always set to 1 in CCMP because an eight-byte header is always needed to
accommodate the large packet number field. The header is shown in Figure 7-7.
6. All the inputs to the CCM encryption engine are available at this point. As input, it takes the
128-bit temporal key, the nonce from step 4, the additional authentication data from step 3,
and the frame body. All the data is authenticated with a 8-byte MIC, and the frame body and
MIC are encrypted. Figure 7-6 illustrates the encryption process.
7. The encrypted frame is prepared for transmission by taking the original MAC header and
appending the CCMP header and the encrypted data from step 6. The resulting frame is pushed
down to the radio interface for transmission, and looks like Figure 7-10.
The encapsulation of a CCMP-protected frame is quite straightforward, and is shown in Figure 7-7.
Following the MAC header, a CCMP header holds the packet number and key ID. The higher-layer
protocol frame and its MIC are encrypted before the FCS.
Figure 7-7. CCMP encapsulation
CCMP reception
When CCMP receives a frame, the encryption and transmission process must be reversed. With no
backwards compatibility baggage requiring the use of the WEP engine, the CCMP decryption process
is a straightforward reversal of Figure 7-6:
1. When a frame is received by the wireless interface and passes the frame check sequence to
ensure it has not been corrupted, it is passed to CCMP for validation.
2. The additional authentication data (AAD) is recovered from the received frame. It consists only
of frame headers, which are transmitted in the clear.
3. The CCMP nonce is also recovered from the frame. It consists of the packet number, the
sender's address, and contents of the quality of service field, all of which are available in the
frame header in the clear.
4. The receiver decrypts the ciphertext. It requires the temporal key, the nonce recovered from
step 3, the authentication data from step 2, and of course, the encrypted frame body. When the
process completes, the receiver has a copy of the decrypted frame plus the decrypted integrity
code.
5. The integrity check is calculated over the plaintext data and the additional authentication data.
If the calculated integrity check value matches the integrity check value extracted in step 4,
then processing proceeds. Otherwise, processing ends.
6. A plaintext frame is constructed from the MAC header and the data recovered in step 4. To pass
through the replay detection check, the packet number must be greater than or equal to the
last received packet number that passed through the integrity check process.
Robust Security Network (RSN) Operations
In addition to defining TKIP and CCMP, 802.11i also defines a set of procedures that build what the
standard calls Robust Security Networks (RSNs). These operations define how keys are derived and
distributed.
802.11i Key Hierarchy
There are two types of keys used by link layer encryption protocols. Pairwise keys protect traffic
between a station and the AP it is currently serving. Group keys protect broadcast or multicast traffic
from an AP to its associated clients. Pairwise keys are ultimately derived from the authentication
information discussed in the previous chapter; group keys are created randomly and distributed to
each station at the whim of the access point.
Pairwise key hierarchy
Both TKIP and CCMP take a single master key and expand it into the different keys required for
frame protection operations. By using key derivation, stations can refresh encryption keys without
re-running the whole authentication process. The master key is the root secret that must be carefully
protected because all keying material is derived from it. Part of the key hierarchy's purpose is to
derive keys used to protect transmission of the temporal keys.
Keying starts with the master key. In the pairwise key hierarchy, which is shown in Figure 7-8, the
master key is unsurprisingly called the pairwise master key (PMK), which is 256 bits long. The PMK
must come from somewhere. In WPA-PSK, the pairwise master key is configured. In configurations
using an authentication server, the master key is computed by the RADIUS server and sent to the
access point in a Microsoft Point-to-Point Encryption (MPPE) vendor-specific RADIUS attribute.
Figure 7-8. Pairwise key hierarchies
To obtain the temporal keys described earlier in this chapter, the PMK is expanded through the use of
a defined pseudorandom function.[*] To further randomize data, the expansion is based on the premaster key, the MAC addresses of both the supplicant and authenticator, and two random nonce
values transmitted as part of the four-way key exchange handshake.
[*]
Many encryption protocols use pseudorandom functions to expand a small seed into a large amount of random data. TLS is
perhaps the best known example.
Both TKIP and CCMP use the pseudorandom function expansion to expand the 256 bits into the
pairwise transient key (PTK). In both the TKIP and CCMP hierarchy, the two chunks of 128 bits of the
transient key are used for keys that protect the temporal keys during distribution.
Both key hierarchies start with two EAPOL keys, used to secure transmission of keying material using
the EAPOL-Key message discussed in the previous chapter. Two 128-bit keys are used. The first, the
EAPOL Key Confirmation Key (KCK), is used to compute message integrity checks on keying
messages. The second, the EAPOL Key Encryption Key (KEK), is used to encrypt keying messages.
Both will be discussed in the section on the four-way handshake.
TKIP's transient key consists of a total of 512 bits, with the additional 256 bits used as the 128-bit
temporal key that is used in TKIP data processing, and the 128-bit key for the Michael integrity
check. TKIP requires two additional keys because it uses traditional encryption and authentication
schemes that strictly separate encryption from authentication. CCMP's transient key is only 384 bits
because only a single 128-bit temporal key is used for authentication and encryption.
Group key hierarchy
Link layer security protocols use a different set of keys for broadcast and multicast transmissions.
Every associated station will have a different pre-master key, and thus, there is no way to derive a
key for use with multiple destinations from the disparate authentication exchanges. Instead, the
authenticator maintains a group master key (GMK) as the basis for temporal keys. The group master
key is expanded into the group key hierarchy shown in Figure 7-9 by the use of a pseudorandom
function. No key encryption or key confirmation keys are generated because the key exchange uses
the pairwise EAPOL keys for key distribution.
Figure 7-9. Group key hierarchy
Networks may update the group keys when stations leave the network, either because they are
finished or are deauthenticated. In TKIP, countermeasures may also trigger the regeneration of the
group keys.
802.11i Key Derivation and Distribution
Rather than simply taking the master secret and using it as the input to a cryptographic protocol,
802.11i specifies a mechanism to derive keys. To prevent replay attacks, the exchange makes use of
random numbers, and requires a handshake. Pairwise and group keys are updated through separate
handshakes, which are both shown in Figure 7-10.
Figure 7-10. Key exchange handshakes
Updating pairwise keys: the four-way handshake
Pairwise, or unicast, keys are distributed through a procedure known as the four-way handshake,
shown in Figure 7-10. Both the supplicant and authenticator are in possession of a shared pairwise
master key. The four-way handshake exchanges parameters used to derive the temporal keys, as
well as confirm that both sides are ready to begin encrypted transmission. Messages in sequence are
implicitly acknowledged by the next message.
1. The authenticator sends the supplicant a nonce, which is a random value that prevents replay
attacks. There is no authentication of the message, but there is no danger from tampering. If
the message is altered, the handshake fails and will be rerun.
At this point, the supplicant can expand the pairwise master key into the full pairwise key
hierarchy. Expansion requires the MAC addresses of the supplicant and authenticator, the
pairwise master key, and the two nonces.
2. The supplicant sends a message that has the supplicant nonce and a copy of the security
2.
parameters from the initial association with the network. The whole message is authenticated
by an integrity check code calculated using the EAPOL Key Confirmation Key.
The authenticator receives the message and extracts the supplicant nonce, which allows the
authenticator to derive the full pairwise key hierarchy. Part of the key hierarchy is the key used
to "sign" the message. If the authenticator cannot validate the message, the handshake fails.
3. Keys are now in place on both sides of the handshake, but need to be confirmed. The
Authenticator sends the supplicant a message indicating the sequence number for which the
pairwise key will be added. It also includes the current group transient key to enable update of
the group key. The group transient key is encrypted using the EAPOL Key Encryption Key, and
the entire message is authenticated using the Key Confirmation Key.
4. The supplicant sends a final confirmation message to the authenticator to indicate that it has
received the keying messages and the authenticator may start using the keys. The message is
authenticated using the Key Confirmation Key.
Updating group keys: the group key handshake
The group key handshake is considerably simpler than the four-way handshake, in part because it
uses part of the results from the four-way handshake. Because the group transient key is encrypted
with the Key Encryption Key from the pairwise hierarchy, the group key handshake requires that a
successful four-way handshake has already occurred. It consists of only two steps:
1. The authenticator sends the group transient key (GTK), encrypted with the Key Encryption Key
from the pairwise key hierarchy. The message is also authenticated with a code calculated with
the Key Confirmation Key.
2. The supplicant sends an acknowledgment message, indicating the authenticator should begin to
use the new key for group frames. This message is also authenticated using the Key
Confirmation Key.
Even though the group key handshake is updating a key used by several stations, the use of the Key
Encryption Key to protect data means that the handshake is inherently pairwise. When the group key
is updated, the group key exchange must be run once for each station.
Although group key updates are generally controlled by the authenticator, stations may request a
group key update by sending an unsolicited confirmation message.
Mixing Encryption Types
To allow for migration between different encryption protocols, as well as to accommodate older
devices incapable of anything stronger than WEP, 802.11i defines a trust hierarchy for encryption
protocols. WEP with 40-bit keys is the weakest protocol, followed by WEP with 104-bit keys, TKIP,
and CCMP.
As part of the initial association to the network, each station can negotiate the encryption protocols it
uses for both unicast and group data. The only restriction is that the group key must use either the
same strength or a weaker encryption protocol. Access points use the "lowest common denominator"
for the group key. In a network where the least capable associated station is only able to run
dynamic WEP, the group key will be dynamic WEP. Other stations may, however, use stronger
unicast protection mechanisms. Many access points provide policy controls to set a minimum
acceptable encryption strength, and may prevent stations from associating with weaker protocols
than desired by the network administrators.
The standard allows for nearly any mixture of encryption methods, with the exception that a station
using CCMP for group frames must only support CCMP for unicast frames. However, many drivers do
not support every allowed mode. Most notably, drivers usually do not support the combination of
CCMP for unicast data in combination with older RC4-based frame encryption for the group key.
Key Caching
Pairwise master keys are the foundation of 802.11i security. Generating them is quite an expensive
operation if the pairwise master key is the result of an 802.1X exchange. Most EAP methods require
multiple messages and a significant amount of computation per step. The 802.1X authentication
process may take several seconds, during which the user is unable to send or receive data. Client
systems located on the boundary between two access points may be particularly affected if the
wireless interface bounces back and forth between two (or more) access points with equivalent signal
strength.
Reducing the authentication overhead is the motivation for PMK caching, which is shown in Figure 711. Rather than require a station to perform the full 802.1X exchange every time it connects to an
access point, it references an existing session by the pairwise master key security association
identifier. If the access point has an existing association, it accepts the association and proceeds
immediately to the four-way handshake. In the four-way handshake, both the supplicant and
authenticator will prove to each other possession of the cached PMK.
Figure 7-11. PMK caching
Stations that do not have cached master keys must perform a full 802.1X authentication to generate
them. One of the motivations for preauthentication, which is discussed in the next chapter, is that
can be used to establish master keys on access points before the handoff occurs, so that a master
key is waiting.
Chapter 8. Management Operations
While being untethered from a wired network can be an advantage, it can lead to problems: the
medium is unreliable, unauthorized users can take advantage of the lack of physical boundaries, and
power consumption is critical when devices are running on batteries. The management features of
the 802.11 protocol were designed to reduce the effect of these problems. In essence, management
operations are all the behind-the-scenes tasks that wireless network devices perform so that the
wireless link looks much like other types of network connections.
802.11 management is a cooperative affair, shared between client devices and the network
infrastructure. Unfortunately, this leaves a great deal of the protocol operation in the hands of
devices that may vary greatly. Some device drivers allow you to customize the management features
discussed in this chapter. Keep in mind, though, that the capabilities of the device driver vary from
one product to another, and the state of wireless networking is such that some vendors are trying to
produce the most feature-rich products possible, while others are aiming at a lower-cost market and
trying to produce the simplest products. The only way to know what is possible is to understand the
capabilities that have been built into the protocol. Then you'll be in a good position to work with
whatever hardware drops in your lap.
Management Architecture
Conceptually, the 802.11 management architecture is composed of three components: the MAC layer
management entity (MLME), a physical-layer management entity (PLME), and a system management
entity (SME). The relation between the different management entities and the related parts of 802.11
is shown in Figure 8-1.
Figure 8-1. Relationship between management entities and components
of the 802.11 specification
802.11 does not formally specify the SME. It is the method by which users and device drivers interact
with the 802.11 network interface and gather information about its status. Both the MAC and PHY
layers have access to a management information base (MIB). The MIB has objects that can be
queried to gain status information, as well as objects that can cause certain actions to take place.
There are three defined interfaces between the management components. The station management
entity may alter both the MAC and PHY MIBs through the MLME and PLME service interfaces.
Additionally, changes to the MAC may require corresponding changes in the PHY, so an additional
interface between the MLME and PLME allows the MAC to make changes to the PHY.
Scanning
Before using any network, you must first find it. With wired networks, finding the network is easy:
look for the cable or a jack on the wall. In the wireless world, stations must identify a compatible
network before joining it. The process of identifying existing networks in the area is called scanning.
Several parameters are used in the scanning procedure. These parameters may be specified by the
user; many implementations have default values for these parameters in the driver.
BSSType (independent, infrastructure, or both)
Scanning can specify whether to seek out independent ad hoc networks, infrastructure
networks, or all networks.
BSSID (individual or broadcast)
The device can scan for a specific network to join (individual) or for any network that is willing
to allow it to join (broadcast). When 802.11 devices are moving, setting the BSSID to
broadcast is a good idea because the scan results will include all BSSs in the area.
SSID ("network name")
The SSID assigns a string of bits to an extended service set. Most products refer to the SSID as
the network name because the string of bits is commonly set to a human-readable string.
Clients wishing to find any network should set this to the broadcast SSID.
ScanType (active or passive)
Active scanning uses the transmission of Probe Request frames to identify networks in the
area. Passive scanning saves battery power by listening for Beacon frames.
ChannelList
Scans must either transmit a Probe Request or listen on a channel for the existence of a
network. 802.11 allows stations to specify a list of channels to try. Products allow configuration
of the channel list in different ways. What exactly constitutes a channel depends on the
physical layer in use. With direct-sequence products, it is a list of channels. With frequencyhopping products, it is a hop pattern.
ProbeDelay
This is the delay, in microseconds, before the procedure to probe a channel in active scanning
begins. This delay ensures that an empty or lightly loaded channel does not completely block
the scan.
MinChannelTime and MaxChannelTime
These values, specified in time units (TUs), specify the minimum and maximum amount of time
that the scan works with any particular channel.
Passive Scanning
Passive scanning saves battery power because it does not require transmitting. In passive scanning,
a station moves to each channel on the channel list and waits for Beacon frames. Any Beacons
received are buffered to extract information about the BSS that sent them.
In the passive scanning procedure, the station sweeps from channel to channel and records
information from any Beacons it receives. Beacons are designed to allow a station to find out
everything it needs to match parameters with the basic service set (BSS) and begin communications.
In Figure 8-2, the mobile station uses a passive scan to find BSSs in its area; it hears Beacon frames
from the first three access points. If it does not hear Beacons from the fourth access point, it reports
that only three BSSs were found.
Figure 8-2. Passive scanning
Active Scanning
In active scanning, a station takes a more assertive role. On each channel, Probe Request frames are
used to solicit responses from a network with a given name. Rather than listening for that network to
announce itself, an active scan attempts to find the network. Stations using active scanning employ
the following procedure for each channel in the channel list:
1.
1. Move to the channel and wait for either an indication of an incoming frame or for the
ProbeDelay timer to expire. If an incoming frame is detected, the channel is in use and can be
probed. The timer prevents an empty channel from blocking the entire procedure; the station
won't wait indefinitely for incoming frames.
2. Gain access to the medium using the basic DCF access procedure and send a Probe Request
frame.
3. Wait for the minimum channel time, MinChannelTime, to elapse.
a. If the medium was never busy, there is no network. Move to the next channel.
b. If the medium was busy during the MinChannelTime interval, wait until the maximum
time, MaxChannelTime, and process any Probe Response frames.
Probe Response frames are generated by networks when they hear a Probe Request that is searching
for the extended service set to which the network belongs. At a party, you might look for a friend by
wandering around the dance floor shouting out her name. (It's not polite, but if you really want to
find your friend, you may not have much choice.) If your friend hears you, she will respondothers will
(you hope) ignore you. Probe Request frames function similarly, but they can also use a broadcast
SSID, which triggers a Probe Response from all 802.11 networks in the area. (It's like shouting
"Fire!" at the partythat's sure to result in a response from everybody!)
One station in each BSS is responsible for responding to Probe Requests. The station that transmitted
the last Beacon frame is also responsible for transmitting any necessary Probe Response frames. In
infrastructure networks, the access points transmit Beacons and thus are also responsible for
responding to itinerant stations searching the area with Probe Requests. IBSSs may pass around the
responsibility of sending Beacon frames, so the station that transmits Probe Response frames may
vary. Probe Responses are unicast management frames and are therefore subject to the positive
acknowledgment requirement of the MAC.
It is common for multiple Probe Responses to be transmitted as a result of a single Probe Request.
The purpose of the scanning procedure is to find every basic service area that the scanning station
can join, so a broadcast Probe Request results in a response from every access point within range.
Any overlapping independent BSSs may also respond.
Figure 8-3 shows the relationship between the transmission of Probe frames and the various timing
intervals that can be configured as part of a scan.
Figure 8-3. Active scanning procedure and medium access
In Figure 8-3 (a), a mobile station transmits a probe request to which two access points respond. The
activity on the medium is shown in Figure 8-3 (b). The scanning station transmits the Probe Request
after gaining access to the medium. Both access points respond with a Probe Response that reports
their network's parameters. Note that the second Probe Response is subject to the rules of the
distributed coordination function and must wait for the contention window to elapse before
transmitting. The first response is transmitted before the minimum response time elapses, so the
station waits until the maximum response time has elapsed before collating the results. In areas with
a large number of networks, it may be necessary to adjust the maximum channel time so the
responses from all the access points in the area can be processed.
Scan Report
A scan report is generated at the conclusion of a scan. The report lists all the BSSs that the scan
discovered and their parameters. The complete parameter list enables the scanning station to join
any of the networks that it discovered. In addition to the BSSID, SSID, and BSSType, the parameters
also include:[*]
[*]
The items actually exposed by any particular software vary.
Beacon interval (integer)
Each BSS can transmit Beacon frames at its own specific interval, measured in TUs.
DTIM period (integer)
DTIM frames are used as part of the powersaving mechanism.
Timing parameters
Two fields assist in synchronizing the station's timer to the timer used by a BSS. The
Timestamp field indicates the value of the timer received by the scanning station; the other
field is an offset to enable a station to match timing information to join a particular BSS.
PHY parameters, CF parameters, and IBSS parameters
These three facets of the network have their own parameter sets, each of which was discussed
in detail in Chapter 4. Channel information is included in the physical-layer parameters.
BSSBasicRateSet
The basic rate set is the list of data rates that must be supported by any station wishing to join
the network. Stations must be able to receive data at all the rates listed in the set. The basic
rate set is composed of the mandatory rates in the Supported Rates information element of
management frames, as in Chapter 4.
What's in a Name? (or, the Security Fallacy of Hidden SSIDs)
WEP is not required by 802.11, and a number of earlier products implement only opensystem authentication. To provide more security than straight open-system
authentication allows, many products offer an "authorized MAC address list." Network
administrators can enter a list of authorized client addresses, and only clients with those
addresses are allowed to connect.
The SSID is an important scanning parameter. Stations search for an SSID when
scanning, and may build a list of SSIDs for presentation to the user. As a unique
identifier for a network, the SSID is often given mythic security properties it does not
actually possess.
At the dawn of 802.11, the SSID was broadcast in the clear in Beacon frames, right
there for the listening. All that was necessary was an 802.11 interface tuned to the right
radio channel. When the stone age of 802.11 began, one vendor began to treat the SSID
as a valuable security token. By enabling the "closed network" option on that vendor's
equipment, the SSID was no longer put in Beacon frames, thus "protecting" the network
from attackers. To further "protect" the SSID from prying eyes, access points operating a
closed network would not respond to Probe Requests with the broadcast SSID.
Closed networks break passive scanning because the SSID is no longer available for easy
collection. In order to prevent a closed network from being completely closed to clients,
however, access points must respond to Probe Requests containing the correct SSID.
Management frames have no encryption, and the SSID value is right there for the taking
in the Probe Request. To be scrupulously correct, the closed network may offer a
vanishingly small incremental amount of security because the SSID is only available
when stations search for the network, rather than several times per second in Beacon
frames.
Hiding an SSID can cause problems with 802.11 management. Although most 802.11
interfaces and their associated drivers can handle hidden SSIDs, not all can. Hiding an
SSID is a nonstandard procedure that can cause problems, and does not provide any real
security. Leave the SSID in the Beacon frames for interoperability, and use a real
security solution like 802.1X if you need it.
Joining
After compiling the scan results, a station can elect to join one of the BSSs. Joining is a precursor to
association; it is analogous to aiming a weapon. It does not enable network access. Before this can
happen, both authentication and association are required.
Choosing which BSS to join is an implementation-specific decision and may even involve user
intervention. BSSs that are part of the same ESS are allowed to make the decision in any way they
choose; common criteria used in the decision are power level and signal strength. Observers cannot
tell when a station has joined a network because the joining process is internal to a node; it involves
matching local parameters to the parameters required by the selected BSS. One of the most
important tasks is to synchronize timing information between the mobile station and the rest of the
network, a process discussed in much more detail in the section "Timer Synchronization," later in this
chapter.
The station must also match the PHY parameters, which guarantees that any transmissions with the
BSS are on the right channel. (Timer synchronization also guarantees that frequency-hopping
stations hop at the correct time, too.) Using the BSSID ensures that transmissions are directed to the
correct set of stations and ignored by stations in another BSS.[*] Capability information is also taken
from the scan result, which matches the use of WEP and any high-rate capabilities. Stations must
also adopt the Beacon interval and DTIM period of the BSS, though these parameters are not as
important as the others for enabling communication.
[*]
Technically, this is true only for stations obeying the filtering rules for received frames. Malicious attackers intent on
compromising network security can easily choose to disobey these rules and capture frames, and most existing product
implementations do not correctly implement the filtering rules.
Authentication
Wireless network media of any sort are open to all sorts of malicious tampering because attackers
can literally immerse themselves in the network medium. Gaining a foothold is trivial if you are
standing in it. Prepare now to enter the authentication twilight zone. There are multiple forms of
"authentication" offered by 802.11, many of which would not pass the laugh test from serious
security experts.
When most network administrators refer to authentication, the implication is that only strong
authentication is worth considering as authentication. Anything that is not based on cryptography
does not prove identity. With the right EAP method, 802.1X authentication can be quite strong, but
802.1X messages can only be exchanged once a system has performed a lower-level 802.11
"authentication" prior to association.
802.11 "Authentication"
802.11 requires that a station establish its identity before sending frames. This initial 802.11
"authentication" occurs every time a station attaches to a network. It should be stressed, however,
that they provide no meaningful network security. There are no cryptographic secrets that are passed
around or validated, and the authentication process is not mutual. It is far more accurate to think of
802.11's low-level authentication as an initial step in the handshake process that a station uses to
attach to the network, and one that identifies the station to the network.
802.11 authentication is a one-way street. Stations wishing to join a network must perform 802.11
authentication to it, but networks are under no obligation to authenticate themselves to a station.
The designers of 802.11 probably felt that access points were part of the network infrastructure and
thus in a more privileged position.
Open-system authentication
Open-system authentication is the only method required by 802.11. Calling it authentication is
stretching the meaning of the term a great deal. In open-system authentication, the access point
accepts the mobile station at face value without verifying its identity. (Imagine a world where similar
authentication applied to bank withdrawals!) Providing network security requires building something
on top of the resulting network session. An open-system authentication exchange consists of two
frames, shown in Figure 8-4.
Figure 8-4. Open-system authentication exchange
The first frame from the mobile station is a management frame of subtype authentication. 802.11
does not formally refer to this frame as an authentication request, but that is its practical purpose. In
802.11, the identity of any station is its MAC address. Like Ethernet networks, MAC addresses must
be unique throughout the network and can readily double as station identifiers. Access points use the
source address of frames as the identity of the sender; no fields within the frame are used to further
identify the sender.
There are two information elements in the body of the authentication request. First, the
Authentication Algorithm Identification is set to 0 to indicate that the open-system method is in use.
Second, the Authentication Transaction Sequence number is set to 1 to indicate that the first frame is
in fact the first frame in the sequence.
The access point then processes the authentication request and returns its response. Like the first
frame, the response frame is a management frame of subtype authentication. Three information
elements are present: the Authentication Algorithm Identification field is set to 0 to indicate opensystem authentication, the Sequence Number is 2, and a Status Code indicates the outcome of the
authentication request. Values for the Status Code are shown in Table 4-6.
Address Filtering (MAC "Authentication")
Roaming is not a word used in the 802.11 standard at all. (A task group recently formed
to address roaming issues, but it is far from completing its work.) However, people use
the word roaming informally a great deal when talking about 802.11. Generally
speaking, most people are referring to the process of moving from one access point to
another.
WEP is not required by 802.11, and a number of earlier products implement only opensystem authentication. To provide more security than straight open-system
authentication allows, many products offer an "authorized MAC address list." Network
administrators can enter a list of authorized client addresses, and only clients with those
addresses are allowed to connect.
While address filtering is better than nothing, it leaves a great deal to be desired. MAC
addresses are generally software-or firmware-programmable and can easily be
overridden by an attacker wishing to gain network access. Distribution of lists of
authorized devices can be painful because it typically must be done for each device in
the network. Furthermore, there is a fair amount of churn on the list. New cards are
purchased, old cards break down, employees leave the organization and take cards with
them, and so on.
Authorized address filtering may be part of a security solution, but it should not be the
linchpin. Rather than depend on MAC address, use 802.1X-based user authentication if
possible. Once network administrators have made the effort to authenticate users,
authentication will be as secure as standards provide for, and address filtering will only
add complexity without any substantial additional security benefit.
The legacy of shared-key authentication
Shared-key authentication makes use of WEP and therefore can be used only on products that
implement WEP, though non-WEP products are now nearly impossible to find. Shared-key
authentication, as its name implies, requires that a shared key be distributed to stations before
attempting authentication. The fundamental theoretical underpinning of shared-key authentication is
that a challenge can be sent to the client, and a response proves possession of the shared key. A
shared-key authentication exchange consists of four management frames of subtype authentication,
shown in Figure 8-5.
The first frame is nearly identical to the first frame in the open-system authentication exchange. Like
the open-system frame, it has information elements to identify the authentication algorithm and the
sequence number; the Authentication Algorithm Identification is set to 1 to indicate shared-key
authentication.
Figure 8-5. Shared-key authentication exchange
Instead of blindly allowing admission to the network, the second frame in a shared-key exchange
serves as a challenge. Up to four information elements may be present in the second frame.
Naturally, the Authentication Algorithm Identification, Sequence Number, and Status Code are
present. The access point may deny an authentication request in the second frame, ending the
transaction. To proceed, however, the Status Code should be set to 0 (success), as shown in Figure
8-5. When the Status Code is successful, the frame also includes a fourth information element, the
Challenge Text. The Challenge Text is composed of 128 bytes generated using the WEP keystream
generator with a random key and initialization vector.
The third frame is the mobile station's response to the challenge. To prove that it is allowed on the
network, the mobile station constructs a management frame with three information elements: the
Authentication Algorithm Identifier, a Sequence Number of 3, and the Challenge Text. Before
transmitting the frame, the mobile station processes the frame with WEP. The header identifying the
frame as an authentication frame is preserved, but the information elements are hidden by WEP.
After receiving the third frame, the access point attempts to decrypt it and verify the WEP integrity
check. If the frame decrypts to the Challenge Text, and the integrity check is verified, the access
point will respond with a status code of successful. Successful decryption of the challenge text proves
that the mobile station has been configured with the WEP key for the network and should be granted
access. If any problems occur, the access point returns an unsuccessful status code.
Defeating shared-key authentication
Shared key authentication is subject to a trivial attack. At the heart of the shared-key authentication
procedure is the encryption of the challenge text. Upon receipt of the random challenge, a station in
posession of the WEP key is able to generate a keystream from the WEP key, then perform an
exclusive-OR operation between the keystream and the data. However, the exclusive-OR is a
reversible operation that relates the challenge text, keystream, and the response text. With any two
items, recovering the third is a matter of reversing the right operation. An attacker can observe the
challenge text and response, and use that to derive a keystream. Although this allows an attacker to
authenticate to a network by playing back the recovered keystream on a new challenge, it does not
allow an attacker to send arbitrary data without recovering the WEP key first. Figure 8-6 illustrates
the keystream recovery procedure.
Figure 8-6. Defeating shared-key authentication
Nevertheless, attacks against shared key authentication are effective enough that it is generally not
recommended. 802.11i does not allow a station that performed 802.11 authentication via shared
keys to become associated to a Robust Security Network.
Preauthentication
Preauthentication is used to speed up association transfer. Authentication can often cause a lag
between the time a station decides to move to a new AP and the time that the frames start flowing
through that AP. Preauthentication attempts to reduce the time by getting the time-consuming
authentication relationship established before it is needed. Due to the overloading of the term
"authentication" by both the low-level 802.11 authentication and the 802.1X authentication, there
are two different types of preauthentication. As it is commonly used by network engineers, though, it
usually refers to the 802.1X authentication.
802.11 Preauthentication
Stations must authenticate with an access point before associating with it, but nothing in 802.11
requires that low-level authentication take place immediately before association. Stations can
802.11-authenticate with several access points during the scanning process so that when association
is required, the station is already authenticated. As a result of preauthentication, stations can
reassociate with access points immediately upon moving into their coverage area, rather than having
to wait for the authentication exchange.
In both parts of Figure 8-7, there is an extended service set composed of two access points. Only one
mobile station is shown for simplicity. Assume the mobile station starts off associated with AP1 at the
left side of the diagram because it was powered on in AP1's coverage area. As the mobile station
moves towards the right, it must eventually associate with AP2 as it leaves AP1's coverage area.
Figure 8-7. Time savings of preauthentication
Preauthentication is not used in the most literal interpretation of 802.11, shown in Figure 8-7 (a). As
the mobile station moves to the right, the signal from AP1 weakens. The station continues monitoring
Beacon frames corresponding to its ESS, and will eventually note the existence of AP2. At some
point, the station may choose to disassociate from AP1, and then authenticate and reassociate with
AP2. These steps are identified in the figure, in which the numbers are the time values from Table 81.
Table 8-1. Chronology for Figure 8-7
Step
Action without preauthentication:
Figure 8-7 (a)
Action with preauthentication: Figure 8-7 (b)
0
Station is associated with AP1
Station is associated with AP1
1
Station moves right into the overlap
between BSS1 and BSS2
Station moves right into the overlap between BSS1
and BSS2 and detects the presence of AP2
1.5
Station preauthenticates to AP2
2
AP2's signal is stronger, so station
decides to move association to AP2
AP2's signal is stronger, so station decides to move
association to AP2
3
Station authenticates to AP2
Station begins using the network
4
Station reassociates with AP2
5
Station begins using the network
Figure 8-7 (b) shows what happens when the station is capable of preauthentication. With this minor
software modification, the station can authenticate to AP2 as soon as it is detected. As the station is
leaving AP1's coverage area, it is authenticated with both AP1 and AP2. The time savings become
apparent when the station leaves the coverage area of AP1: it can immediately reassociate with AP2
because it is already authenticated. Preauthentication makes roaming a smoother operation because
authentication can take place before it is needed to support an association. All the steps in Figure 8-7
(b) are identified by time values from Table 8-1.
802.11i Preauthentication and Key Caching
When a network is authenticated with 802.1X, the most time-consuming step in getting from the
802.11 join to the ability to send network protocol packets is the 802.1X authentication, especially if
it uses an EAP method with several frame round-trips. Preauthentication, shown in Figure 8-8, allows
a station to establish a security context with a new AP before associating to it. In essence,
preauthentication decouples the association and security procedures, and allows them to be
performed independently. WPA explicitly excluded preauthentication.
Figure 8-8. 802.11i preauthentication
Figure 8-8 shows the following sequence of steps.
1. The station associates to the first access point it finds on the network. It selects this AP based
on the criteria in its firmware.
2. Once associated, the station can perform an 802.1X authentication. This step uses EAPOL
frames as described in Chapter 6, with an Ethertype of (hexadecimal) 88-8E. EAPOL frames are
converted into RADIUS packets by the AP, and the session is authenticated.
3. Dynamic keys for the radio link are derived on both sides through the four-way handshake for
pairwise keys and the group key handshake for the group keys.
4. With keys configured, the station is "on the air" and can send and receive network protocol
packets.
Station software is in control of roaming behavior, and can use that to its advantage. As the
station moves in such a way that AP2 appears to be a better choice, it can perform
preauthentication to speed up the process of moving over to AP2. Rather than move everything
all at once, though, it performs preauthentication to cut down on the interruption between
sending network packets.
5. Preauthentication commences with an EAPOL-Start message sent from the station to the new
AP. A station can only be associated with a single AP, so the preauthentication frames are
channeled through the old AP. Preauthentication is a complete 802.1X exchange.
a. Preauthentication frames use the Ethertype of (hexadecimal) 88-C7 because most APs
apply special processing to the regular authentication Ethertype. The source address of the
frames is the station, the receiver address is the BSSID of its current AP (in this case, the
MAC address of AP1's wireless interface), and the destination address is the BSSID of the
new AP (in this case, AP2's wireless interface).
b. When received by AP1, the frames are sent over the distribution system to AP2. The AP
only has a MAC address. If the two APs are not connected directly to the same Ethernet
broadcast domain, they must have an alternative method of shuttling preauthentication
frames between devices.
c.
c. During this entire step, the station remains associated to AP1, and can send and receive
network packets through its existing encrypted connection. Because the station is still on
the air, there is no apparent authentication occuring.
d. The result of the preauthentication is that a security context with AP2 is established. The
station and AP2 have derived a pairwise master key, which can be further processed to
create keys between the station and AP2. Both the station and the AP store the pairwise
master key in a key cache.
6. When the station pulls the trigger, the association is moved to AP2. As part of the initial
association, the station includes a copy of its key cache to tell AP2 that it was already
authenticated.
7. AP2 receives the authentication request and searches its key cache. Finding an entry, it starts
the fourway pairwise key handshake immediately. By proceeding to key derivation, the station
is unable to send and receive packets for only a short time.
802.11 preauthentication moves the time-consuming 802.1X EAP method to occur in parallel with
sending and receiving network frames on an authenticated connection. The first association will be
slow because the full EAP exchange is required. On subsequent associations, however,
preauthentication can dramatically reduce handoff times.
Association
Once authentication has completed, stations can associate with an access point (or reassociate with a
new access point) to gain full access to the network. Association is a recordkeeping procedure that
allows the distribution system to track the location of each mobile station, so frames destined for the
mobile station can be forwarded to the correct access point. After association completes, an access
point must register the mobile station on the network so frames for the mobile station are delivered
to the access point. One method of registering is to send a gratuitous ARP so the station's MAC
address is associated with the switch port connected to the access point.
Association is restricted to infrastructure networks and is logically equivalent to plugging into a wired
network. Once the procedure is complete, a wireless station can use the distribution system to reach
out to the world, and the world can respond through the distribution system. 802.11 explicitly forbids
associating with more than one access point.
Association Procedure
The basic association procedure is shown in Figure 8-9.
Figure 8-9. Association procedure
Like authentication, association is initiated by the mobile station. No sequence numbers are needed
because the association process is a three-step exchange. The two frames are management frame
subtypes defined by the specification. As unicast management frames, both steps in the association
procedure are composed of an association frame and the required link-layer acknowledgment:
1. Once a mobile station has authenticated to an access point, it can issue an Association Request
frame. Stations that have not yet authenticated receive a Deauthentication frame from the
access point in response.
2. The access point then processes the association request. 802.11 does not specify how to
determine whether an association should be granted; it is specific to the access point
implementation. One common consideration is the amount of space required for frame
buffering. Rough estimates are possible based on the Listen Interval in the Association Request
frame.
a.
a. When the association request is granted, the access point responds with a status code of 0
(successful) and the Association ID (AID). The AID is a numerical identifier used to
logically identify the mobile station to which buffered frames need to be delivered. More
detail on the process can be found in the "Power Conservation" section of this chapter.
b. Unsuccessful association requests include only a status code, and the procedure ends.
3. The access point begins processing frames for the mobile station. In all commonly used
products, the distribution system medium is Ethernet. When an access point receives a frame
destined for an associated mobile station, that frame can be bridged from the Ethernet to the
wireless medium or buffered if the mobile station is in a powersaving state. In shared
Ethernets, the frame will be sent to all the access points and will be bridged by the correct one.
In switched Ethernets, the station's MAC address will be associated with a particular switch port.
That switch port is, of course, connected to the access point currently providing service for the
station.
Reassociation Procedure
Reassociation is the process of moving an association from an old access point to a new one. Over
the air, it is almost the same as an association; on the backbone network, however, access points
may interact with each other to move frames. When a station moves from the coverage area of one
access point to another, it uses the reassociation process to inform the 802.11 network of its new
location. The procedure is shown in Figure 8-10.
Figure 8-10. Reassociation procedure
The mobile station begins the procedure associated with an access point. The station monitors the
quality of the signal it receives from that access point, as well as the signal quality from other access
points in the same ESS. When the mobile station detects that another access point would be a better
choice, it initiates the reassociation procedure. The factors used to make that decision are productdependent. Received signal strength can be used on a frame-by-frame basis, and the constant
Beacon transmissions provide a good baseline for signal strength from an access point. Before the
first step, the mobile station must authenticate to the new access point if it has not done so already.
Figure 8-10 depicts the following steps:
1. The mobile station issues a Reassociation Request to the new access point. Reassociation
Requests have content similar to Association Requests. The only difference is that Reassociation
Request frames contain a field with the address of the old access point. The new access point
must communicate with the old access point to determine that a previous association did exist.
Although a standard inter-access point protocol was defined by the IEEE, many implementations
remain proprietary. If the new access point cannot verify that the old access point authenticated
the station, the new access point responds with a Deauthentication frame and ends the
procedure.
2. The access point processes the Reassociation Request. Processing Reassociation Requests is
similar to processing Association Requests; the same factors may be used in deciding whether
to allow the reassociation:
a. If the Reassociation Request is granted, the access point responds with a Status Code of 0
(successful) and the AID.
b. Unsuccessful Reassociation Requests include just a Status Code, and the procedure ends.
3. The new access point contacts the old access point to finish the reassociation procedure. This
communication is part of the IAPP.
4. The old access point sends any buffered frames for the mobile station to the new access point.
802.11 does not specify the communication between access points. At the conclusion of the
buffered frame transfer:
a. Any frames buffered at the old access point are transferred to the new access point so
they can be delivered to the mobile station.
b. The old access point terminates its association with the mobile station. Mobile stations are
allowed to associate with only one access point at any given time.
5. The new access point begins processing frames for the mobile station. When it receives a frame
destined for the mobile station, that frame is bridged from the Ethernet to the wireless medium
or buffered for a mobile station in a powersaving mode.
Reassociation is also used to rejoin a network if the station leaves the coverage area and returns
later to the same access point. Figure 8-11 illustrates this scenario.
Figure 8-11. Reassociation with the same access point
What Is Roaming?
Roaming is not a word used in the 802.11 standard at all. (A task group recently formed
to address roaming issues, but it is far from completing its work.) However, people use
the word roaming informally a great deal when talking about 802.11. Generally
speaking, most people are referring to the process of moving from one access point to
another.
Roaming has suffered a bit from "buzzword overload" in the past few years, and now
means different things to different speakers. At the most basic level, roaming is the
process of moving a station between APs. How does a station decide to move between
APs? The 802.11 standard has nothing to say on the matter. Decisions to move between
APs are based entirely on the hardware and software, and up to each manufacturer.
Some client devices will pick up on the strongest signal available when the interface is
initialized, and hang on for dear life, only moving when the initial AP is no longer
reachable. Some devices will always head for the strongest available signal, which may
result in flip-flops between APs when two signals are evenly balanced. Still other devices
will incorporate recent history into roaming decisions to avoid excessive movements
between two nearby APs. How stations decide if and when to move between APs is
entirely up to the manufacturer. With apologies to Milton Friedman, roaming is always
and everywhere a client phenomenon.
Digging deeper into a second level of meaning, "roaming" is also used to discuss how a
station can change APs while keeping active network connections open. If IP subnet
boundaries are involved, this is an added level of complexity that needs to be addressed
outside any standards. Chapter 21 discusses how to build a wireless LAN, and discusses
how to achieve seamless roaming across arbitrary network topologies.
Roaming is sometimes given yet another meaning, referring to the process of moving a
network session from one network (say, an 802.11 wireless LAN) to another disparate
network (say, a 3G mobile telephone network). The IEEE has started another working
group to define roaming operations to move network sessions and state between
different types of IEEE 802 networks.
Power Conservation
The major advantage of wireless networks is that network access does not require nodes to be in any
particular location. To take full advantage of mobility, nothing can constrain the location of a node,
including the availability of electrical power. Mobility therefore implies that most mobile devices can
run on batteries. But battery power is a scarce resource; batteries can run only so long before they
need to be recharged. Requiring mobile users to return frequently to commercial power is
inconvenient, to say the least. Many wireless applications require long battery life without sacrificing
network connectivity.
As with any other network interface, powering down the transceiver can lead to great power savings
in wireless networks. When the transceiver is off, it is said to be sleeping, dozing, or in powersaving
mode (PS). When the transceiver is on, it is said to be awake, active, or simply on. Power
conservation in 802.11 is achieved by minimizing the time spent in the latter stage and maximizing
the time in the former. 802.11 accomplishes this without sacrificing connectivity.
Power Management in Infrastructure Networks
Power management can achieve the greatest savings in infrastructure networks. All traffic for mobile
stations must go through access points, so they are an ideal location to buffer traffic. Most traffic can
be buffered. The standard forbids buffering frames that require in-order delivery and have Order bit
set because buffer implementations could possibly re-order frames. There is no need to work on a
distributed buffer system that must be implemented on every station; the bulk of the work is left to
the access point. By definition, access points are aware of the location of mobile stations, and a
mobile station can communicate its power management state to its access point. Furthermore,
access points must remain active at all times; it is assumed that they have access to continuous
power. Combining these two facts allows access points to play a key role in power management on
infrastructure networks.
Access points have two power management-related tasks. First, because an access point knows the
power management state of every station that has associated with it, it can determine whether a
frame should be delivered to the wireless network because the station is active or buffered because
the station is asleep. But buffering frames alone does not enable mobile stations to pick up the data
waiting for them. An access point's second task is to announce periodically which stations have
frames waiting for them. The periodic announcement of buffer status also helps to contribute to the
power savings in infrastructure networks. Powering up a receiver to listen to the buffer status
requires far less power than periodically transmitting polling frames. Stations only need to power up
the transmitter to transmit polling frames after being informed that there is a reason to expend the
energy.
Power management is designed around the needs of the battery-powered mobile stations. Mobile
stations can sleep for extended periods to avoid using the wireless network interface. Part of the
association request is the Listen Interval parameter, the number of Beacon periods for which the
mobile station may choose to sleep. Longer listen intervals require more buffer space on the access
point; therefore, the Listen Interval is one of the key parameters used in estimating the resources
required to support an association. The Listen Interval is a contract with the access point. In agreeing
to buffer any frames while the mobile station is sleeping, the access point agrees to wait for at least
the listen interval before discarding frames. If a mobile station fails to check for waiting frames after
each listen interval, they may be discarded without notification.
Unicast frame buffering and delivery using the Traffic Indication Map (TIM)
When frames are buffered, the destination node's Association ID (AID) provides the logical link
between the frame and its destination. Each AID is logically connected to frames buffered for the
mobile station that is assigned that AID. Multicast and broadcast frames are buffered and linked to
an AID of zero. Delivery of buffered multicast and broadcast frames is treated in the next section.
Buffering is only half the battle. If stations never pick up their buffered frames, saving the frames is
a rather pointless exercise. To inform stations that frames are buffered, access points periodically
assemble a traffic indication map (TIM) and transmit it in Beacon frames. The TIM is a virtual bitmap
composed of 2,008 bits; offsets are used so that the access point needs to transmit only a small
portion of the virtual bitmap. This conserves network capacity when only a few stations have buffered
data. Each bit in the TIM corresponds to a particular AID; setting the bit indicates that the access
point has buffered unicast frames for the station with the AID corresponding to the bit position.
Mobile stations must wake up and enter the active mode to listen for Beacon frames to receive the
TIM. By examining the TIM, a station can determine if the access point has buffered traffic on its
behalf. To retrieve buffered frames, mobile stations use PS-Poll Control frames. When multiple
stations have buffered frames, all stations with buffered data must use the random backoff algorithm
before transmitting the PS-Poll.
Each PS-Poll frame is used to retrieve one buffered frame. That frame must be positively
acknowledged before it is removed from the buffer. Positive acknowledgment is required to keep a
second, retried PS-Poll from acting as an implicit acknowledgment. Figure 8-12 illustrates the
process.
If multiple frames are buffered for a mobile station, then the More Data bit in the Frame Control field
is set to 1. Mobile stations can then issue additional PS-Poll requests to the access point until the
More Data bit is set to 0, though no time constraint is imposed by the standard.
Figure 8-12. PS-Poll frame retrieval
After transmitting the PS-Poll, a mobile station must remain awake until either the polling transaction
has concluded or the bit corresponding to its AID is no longer set in the TIM. The reason for the first
case is obvious: the mobile station has successfully polled the access point; part of that transaction
was a notification that the mobile station will be returning to a sleeping mode. The second case
allows the mobile station to return to a power conservation mode if the access point discards the
buffered frame. Once all the traffic buffered for a station is delivered or discarded, the station can
resume sleeping.
The buffering and delivery process is illustrated in Figure 8-13, which shows the medium as it
appears to an access point and two associated powersaving stations. The hash marks on the timeline
represent the beacon interval. Every beacon interval, the access point transmits a Beacon frame with
a TIM information element. (This figure is somewhat simplified. A special kind of TIM is used to
deliver multicast traffic; it will be described in the next section.) Station 1 has a listen interval of 2,
so it must wake up to receive every other TIM, while station 2 has a listen interval of 3, so it wakes
up to process every third TIM. The lines above the station base lines indicate the ramp-up process of
the receiver to listen for the TIM.
At the first beacon interval, there are frames buffered for station 1. No frames are buffered for
station 2, though, so it can immediately return to sleep. At the second beacon interval, the TIM
indicates that there are buffered frames for stations 1 and 2, although only station 1 woke up to
listen to the TIM. Station 1 issues a PS-Poll and receives the frame in response. At the conclusion of
the exchange, station 1 returns to sleep. Both stations are asleep during the third beacon. At the
fourth beacon, both wake up to listen to the TIM, which indicates that there are frames buffered for
both.
Figure 8-13. Buffered frame retrieval process
Both station 1 and station 2 prepare to transmit PS-Poll frames after the expiration of a contention
window countdown, as described in Chapter 3. Station 1 wins because its random delay was shorter.
Station 1 issues a PS-Poll and receives its buffered frame in response. During the transmission,
station 2 defers. If, at the end of that frame transmission, a third station, which is not illustrated,
seizes the medium for transmission, station 2 must continue to stay awake until the next TIM. If the
access point has run out of buffer space and has discarded the buffered frame for station 2, the TIM
at the fifth beacon indicates that no frames are buffered, and station 2 can finally return to a lowpower mode.
Stations may switch from a power conservation mode to active mode at any time. It is common for
laptop computers to operate with full power to all peripherals when connected to AC power and
conserve power only when using the battery. If a mobile station switches to the active mode from a
sleeping mode, frames can be transmitted without waiting for a PS-Poll. PS-Poll frames indicate that
a powersaving mobile station has temporarily switched to an active mode and is ready to receive a
buffered frame. By definition, active stations have transceivers operating continuously. After a switch
to active mode, the access point can assume that the receiver is operational, even without receiving
explicit notification to that effect.
Access points must retain frames long enough for mobile stations to pick them up, but buffer
memory is a finite resource. 802.11 mandates that access points use an aging function to determine
when buffered frames are old enough to be discarded. The standard leaves a great deal to the
discretion of the developer because it specifies only one constraint. Mobile stations depend on access
points to buffer traffic for at least the listen interval specified with the association, and the standard
forbids the aging function from discarding frames before the listen interval has elapsed. Beyond that,
however, there is a great deal of latitude for vendors to develop different buffer management
routines.
Delivering multicast and broadcast frames: the Delivery TIM (DTIM)
Frames with a group address cannot be delivered using a polling algorithm because they are, by
definition, addressed to a group. Therefore, 802.11 incorporates a mechanism for buffering and
delivering broadcast and multicast frames. Buffering is identical to the unicast case, except that
frames are buffered whenever any station associated with the access point is sleeping. Buffered
broadcast and multicast frames are saved using AID 0. Access points indicate whether any broadcast
or multicast frames are buffered by setting the first bit in the TIM; this bit corresponds to AID 0.
Each BSS has a parameter called the DTIM Period. TIMs are transmitted with every Beacon. At a
fixed number of Beacon intervals, a special type of TIM, a Delivery Traffic Indication Map (DTIM), is
sent. The TIM element in Beacon frames contains a counter that counts down to the next DTIM; this
counter is zero in a DTIM frame. Buffered broadcast and multicast traffic is transmitted after a DTIM
Beacon. Multiple buffered frames are transmitted in sequence; the More Data bit in the Frame
Control field indicates that more frames must be transmitted. Normal channel acquisition rules apply
to the transmission of buffered frames. The access point may choose to defer the processing of
incoming PS-Poll frames until the frames in the broadcast and multicast transmission buffers have
been transmitted.
Figure 8-14 shows an access point and one associated station. The DTIM interval of the access point
is set to 3, so every third TIM is a DTIM. Station 1 is operating in a sleep mode with a listen interval
of 3. It will wake up on every third beacon to receive buffered broadcast and multicast frames. After
a DTIM frame is transmitted, the buffered broadcast and multicast frames are transmitted, followed
by any PS-Poll exchanges with associated stations. At the second beacon interval, only broadcast and
multicast frames are present in the buffer, and they are transmitted to the BSS. At the fifth beacon
interval, a frame has also been buffered for station 1. It can monitor the map in the DTIM and send a
PS-Poll after the transmission of buffered broadcast and multicast frames has concluded.
Figure 8-14. Multicast and broadcast buffer transmission after DTIMs
To receive broadcast and multicast frames, a mobile station must be awake for DTIM transmissions.
Nothing in the specification, however, keeps powersaving stations in infrastructure networks from
waking up to listen to DTIM frames. Some products that implement powersaving modes will attempt
to align their awakenings with DTIM transmissions. If the system administrator determines that
battery life is more important than receiving broadcast and multicast frames, a station can be
configured to sleep for its listen period without regard to DTIM transmissions. Some documentation
may refer to this as extremely low power, ultra powersaving mode, deep sleep, or something similar.
Several products allow configuration of the DTIM interval. Lengthening the DTIM interval allows
mobile stations to sleep for longer periods and maximizes battery life at the expense of timely
delivery. Shorter DTIM intervals emphasize quick delivery at the expense of more frequent power-up
and power-down cycles. You can use a longer DTIM when battery life is at a premium and delivery of
broadcast and multicast frames is not important. Whether this is appropriate depends on the
applications you are using and how they react to long link-layer delays.
IBSS Power Management
Power management in an IBSS is not as efficient as power management in an infrastructure network.
In an IBSS, far more of the burden is placed on the sender to ensure that the receiver is active.
Receivers must also be more available and cannot sleep for the same lengths of time as in
infrastructure networks.
As in infrastructure networks, power management in independent networks is based on traffic
indication messages. Independent networks must use a distributed system because there is no
logical central coordinator. Stations in an independent network use announcement traffic indication
messages (ATIMs), which are sometimes called ad hoc traffic indication messages, to preempt other
stations from sleeping. All stations in an IBSS listen for ATIM frames during specified periods after
Beacon transmissions.
If a station has buffered data for another station, it can send an ATIM frame as notification. In effect,
the ATIM frame is a message to keep the transceiver on because there is pending data. Stations that
do not receive ATIM frames are free to conserve power. In Figure 8-15 (a), station A has buffered a
frame for station C, so it sends a unicast ATIM frame to station C during the ATIM transmission
window, which has the effect of notifying station C that it should not enter powersaving mode.
Station B, however, is free to power down its wireless interface. Figure 8-15 (b) shows a multicast
ATIM frame in use. This frame can be used to notify an entire group of stations to avoid entering lowpower modes.
Figure 8-15. ATIM usage
A time window called the ATIM window follows the Beacon transmission. This window is the period
during which nodes must remain active. No stations are permitted to power down their wireless
interfaces during the ATIM window. It starts at the time when the beacon is expected and ends after
a period specified when the IBSS is created. If the beacon is delayed due to a traffic overrun, the
usable portion of the ATIM window shrinks by the same amount.
The ATIM window is the only IBSS-specific parameter required to create an IBSS. Setting it to 0
avoids using any power management. Figure 8-16 illustrates the ATIM window and its relation to the
beacon interval. In the figure, the fourth beacon is delayed due to a busy medium. The ATIM window
remains constant, starting at the target beacon interval and extending the length of the ATIM
window. The usable period of the ATIM window shrinks by the length of the delay in beacon
transmission.
Figure 8-16. ATIM window
To monitor the entire ATIM window, stations must wake up before the target beacon transmission.
Four situations are possible: the station has transmitted an ATIM, received an ATIM, neither
transmitted nor received, or both transmitted and received. Stations that transmit ATIM frames must
not sleep. Transmitting an ATIM indicates an intent to transmit buffered traffic and thus an intent to
stay active. Stations to which ATIM frames are addressed must also avoid sleeping so they can
receive any frames transmitted by the ATIM's sender. If a station both transmits and receives ATIM
frames, it stays up. A station is permitted to sleep only if it neither transmits nor receives an ATIM.
When a station stays up due to ATIM traffic, it remains active until the conclusion of the next ATIM
window, as shown in Figure 8-17. In the figure, the station goes active for the first ATIM window. If it
does not send or receive any ATIM frames, it sleeps at the end of the ATIM window. If it sends or
receives an ATIM frame, as in the second ATIM window, the station stays active until the conclusion
of the third ATIM window.
Figure 8-17. ATIM effects on powersaving modes
Only certain control and management frames can be transmitted during the ATIM window: Beacons,
RTS, CTS, ACK, and, of course, ATIM frames. Transmission takes place according to the rules of the
DCF. ATIM frames may be transmitted only during the ATIM window because stations may be
sleeping outside the ATIM window. Sending an ATIM frame is useless if other stations in the IBSS are
sleeping. In the same vein, acknowledgments are required for unicast ATIM frames because that is
the only guarantee that the ATIM was received and that the frame destination will be active for the
remainder of the beacon interval. Acknowledgments are not required for multicast ATIM frames
because multicast frames cannot be efficiently acknowledged by a large group of stations. If all
potential recipients of an ATIM frame were required to acknowledge it, the mass of acknowledgments
could potentially interrupt network service.
Buffered broadcast and multicast frames are transmitted after the conclusion of the ATIM window,
subject to DCF constraints. Following the transmission of broadcast and multicast frames, a station
may attempt to transmit unicast frames that were announced with an ATIM and for which an
acknowledgment was received. Following all transmissions announced with an ATIM, stations may
transmit unbuffered frames to other stations that are known to be active. Stations are active if they
have transmitted the Beacon, an ATIM, or are not capable of sleeping. If contention is severe enough
to prevent a station from sending the buffered frame it announced with an ATIM, the station must
reannounce the transmission with an ATIM at the start of the next ATIM window.
Figure 8-18 illustrates several of these rules. In the first beacon interval, the first station transmits a
multicast ATIM to stations 2, 3, and 4. Multicast ATIM frames need not be acknowledged, but the
transmission of the ATIM means that all stations must remain active for the duration of the first
beacon window to receive multicast frames from station 1. When the ATIM window ends, station 1
can transmit its multicast frame to the other three stations. After doing so, station 4 can take
advantage of the remaining time before the beacon to transmit a frame to station 1. It was not
cleared with an ATIM, but it is known to be active.
Figure 8-18. Effect of ATIM on powersaving modes in an IBSS network
In the second beacon interval, stations 2 and 3 have both buffered a frame for station 4, so each
transmits an ATIM. Station 4 acknowledges both. At the conclusion of the ATIM window, station 1 has
neither transmitted nor received an ATIM and can enter a low-power state until the next beacon
interval. However, station 2's frame is extremely long and robs station 3 of the opportunity to
transmit its frame.
Station 3 still has a buffered frame for station 4 when the third beacon interval opens. It therefore
retransmits its ATIM frame to station 4, which is acknowledged. Station 2 is not involved in any ATIM
exchanges and can enter a low-power state when the ATIM window ends. At that time, no broadcast
or multicast frames have been buffered, and the ATIM-cleared frame from station 3 to station 4 can
be transmitted. After the frame from 3 to 4 is transmitted, station 4 can again take advantage of the
remaining time before the beacon frame to transmit a frame of its own to station 3, which is known
to be active because of the ATIM exchange.
Stations are responsible for maintaining sufficient memory to buffer frames, but the buffer size must
be traded off against the use of that memory for other purposes. The standard allows a station in an
independent network to discard frames that have been buffered for an "excessive" amount of time,
but the algorithm used to make that determination is beyond the scope of the standard. The only
requirement placed on any buffer management function is that it retain frames for at least one
beacon period.
Timer Synchronization
Like other wireless network technologies, 802.11 depends a great deal on the distribution of timing
information to all the nodes. It is especially important in frequency-hopping networks because all
stations on the network must change frequency channels in a coordinated pattern. Timing
information is also used by the medium reservation mechanisms.
In addition to local station timing, each station in a basic service area maintains a copy of the timing
synchronization function (TSF), which is a local timer synchronized with the TSF of every other
station in the basic service area. The TSF is based on a 1-MHz clock and "ticks" in microseconds.
Beacon frames are used to periodically announce the value of the TSF to other stations in the
network. The "now" in a timestamp is when the first bit of the timestamp hits the PHY for
transmission.
Infrastructure Timing Synchronization
The ease of power management in an infrastructure network is based on the use of access points as
central coordinators for data distribution and power management functions. Timing in infrastructure
networks is quite similar. Access points are responsible for maintaining the TSF time, and any
stations associated with an access point must simply accept the access point's TSF as valid.
When access points prepare to transmit a Beacon frame, the access point timer is copied into the
Beacon's timestamp field. Stations associated with an access point accept the timing value in any
received Beacons, but they may add a small offset to the received timing value to account for local
processing by the antenna and transceiver. Associated stations maintain local TSF timers so they can
miss a Beacon frame and still remain roughly synchronized with the global TSF. The wireless medium
is expected to be noisy, and Beacon frames are unacknowledged. Therefore, missing a Beacon here
and there is to be expected, and the local TSF timer mitigates against the occasional loss of Beacon
frames.
To assist active scanning stations in matching parameters with the BSS, timing values are also
distributed in Probe Response frames. When a station finds a network by scanning, it saves the
timestamp from the Beacon or Probe Response and the value of the local timer when it was received.
To match the local timer to the network timer, a station then takes the timestamp in the received
network advertisement and adds the number of microseconds since it was received. Figure 8-19
illustrates this process.
Figure 8-19. Matching the local timer to a network timer
IBSS Timing Synchronization
IBSSs lack a central coordination point, so the Beacon process is distributed. TSF maintenance is a
subset of the Beacon generation process. Time is divided into segments equivalent to the interbeacon
timing period. Beacon frames are supposed to be transmitted exactly as the beacon interval ends, at
the so-called target Beacon transmission time (TBTT). Independent networks take the TBTT as a
guideline.
All stations in the IBSS prepare to transmit a Beacon frame at the target time. As it approaches, all
other traffic is suspended. Timers for the transmission of frames other than Beacon frames or ATIM
frames are stopped and held to clear the medium for the important management traffic. All stations
in the IBSS generate a backoff timer for Beacon transmission; the backoff timer is a random delay
between zero and twice the minimum contention window for the medium. After the target beacon
interval, all stations begin to count the Beacon backoff timer down to zero. If a Beacon is received
before the station's transmission time, the pending Beacon transmission is canceled.
In Figure 8-20, each station selects a random delay; station 2 has randomly generated the shortest
delay. When station 2's timer expires, it transmits a Beacon, which is received by stations 1 and 3.
Both stations 1 and 3 cancel their Beacon transmissions as a result. Because timer synchronization
ensures that all stations have synchronized timers, multiple Beacon frames do not pose a problem.
Receivers simply process multiple Beacon frames and perform multiple updates to the TSF timer.
Beacon generation interacts closely with power management. Beacon frames must be generated
during the active period around each Beacon interval so that all stations are available to process the
Beacon. Furthermore, the Beacon sender is not allowed to enter a low-power state until the end of
the next active period. The latter rule ensures that at least one station is awake and can respond to
probes from new stations scanning to discover networks.
Rules for adopting the received timestamp are more complex in an independent network. No
centralized timer exists, so the goal of the standard is to synchronize all timers to the timer of the
fastest-running clock in the BSS. When a Beacon is received, the
Figure 8-20. Distributed Beacon generation
timestamp is adjusted for processing delays and compared to the local TSF. The received timestamp
updates the local timer only if it is later than the local timer.
Spectrum Management
802.11a was originally developed as a standard for the United States only. European radio
regulations for the 5 GHz frequency band are stricter, which necessitated the development of special
procedures to adapt the 802.11 MAC for use in Europe, which were eventually standardized as
802.11h in 2003.
Transmit Power Control (TPC)
Transmit power control is required by European regulations[*] to ensure that 5 GHz radio
transmitters stay within regulatory power limits and avoid interfering with certain satellite services.
Developing better control over transmission power brings many other benefits, some of which have
been long known in the world of mobile telephony. High-powered client transmissions may cover very
large areas. In a densely-deployed network of APs, a single client at high power may have much
higher range than is necessary. Long range is not always a plus. Running a radio transmitter at high
power decreases battery life. If the range of a transmission is longer than necessary, the extra reach
represents "wasted" power. Higher power may also lead to a reduction in network throughput as
client devices interfere with each other unnecessarily.
[*]
ERC/DEC/(99)23, available at http://www.ero.dk/doc98/Official/Pdf/DEC9923E.PDF.
Figure 8-21. High client power interference with multiple APs
For an illustration of problems caused by high client power, consider Figure 8-21. In the figure, nine
APs are shown in a network.[*] Client #1 is associated with the center AP in the top row. If the client
computer is configured to transmit at maximum power, the range of the transmission will reach out
to the outer circle. However, the transmission to the AP requires power to only the inner circle. The
difference between the two circles is the excess range caused by transmitting too high. Running the
radio transmitter higher than necessary will drain the battery on a portable device faster than is
strictly necessary. (To extend the battery life of phones, mobile telephone networks have
incorporated transmit power limitation techniques for many years.)
[*]
Note that laying out a nine-AP network using only three non-overlapping channels is not possible, which is a strong argument
for using 802.11a in high-density environments.
Controlling the range of transmission also makes the network function more smoothly. All
transmissions must gain exclusive access to the radio medium. When excessive power is used, the
area covered by a transmission is much larger than it should be. Any station operating on channel 11
within the outer circle must defer to a frame in progress from Client #1. The shaded area between
the optimum transmit power and the excessive power represents the area which is blocked because
Client #1 is transmitting at power that is too high. For example, Client #2 is blocked from
communicating with its nearby AP because it must share the radio medium. Reducing transmission
power to the level required to reach only the serving AP may improve communications throughout
the network by limiting overlap between nearby APs.
Basic operation of transmit power control
Transmit power control (TPC) is an 802.11 service that attempts to hold transmit power to the lowest
possible productive level. It takes into account the maximum power allowed by regulators, as well as
further constraints. Although designed to satisfy regulatory requirements, transmit power control
may have additional benefits.
The absolute cap on any radio transmission is set by regulatory authorities, and is found in the
relevant documents and rules published by the authority. Regulatory maximum power may be
configured into an AP or station, or it may be learned from Beacon frames containing Country
elements. Radio transmissions may be subject to additional constraints that further reduce the
maximum power. European regulations specify a further mitigation requirement of at least 3 dB to
reduce interference with satellite services.
Maximum transmission power is specified using the Country element in Beacon frames, and is
available to any station wishing to associate to a network. The Country element specifies the
regulatory maximum power, and the Power Constraint element can be used to specify a lower
maximum transmission power specific to the network.
Before beginning operation, stations must calculate the maximum transmission power that may be
used. Typically, this is calculated by taking the regulatory maximum power and subtracting any
constraints for mitigation or additional local constraints. For example, network administrators may
wish to specify a restrictive local maximum transmission power to reduce range and therefore
interference.
Changes to the association process
When a spectrum management-capable station associates (or reassociates) to an access point, it
must supply the minimum and maximum transmission power in a Power Capability information
element. Access points may incorporate the information supplied by a client into the association
process, and are free to use it any manner. How, or even whether, an access point uses the power
capability information is not specified by 802.11 standards. Stations that have high power and may
violate regulations may be rejected to preserve regulatory compliance. Access points may also reject
clients with low transmit power capability; the standard suggests that it may increase the possibility
of hidden nodes, although that seems unlikely.
Changing the transmission power
Both access points and client stations may dynamically adjust the transmission power on a frame-byframe basis. For each frame, the receiver may compute the link margin, the amount by which the
received power exceeds the minimum acceptable value. Link margin is the margin of safety. If a
station's transmissions were received at the minimum acceptable value, the link margin would be
zero, indicating that any detrimental change to the link would interrupt communication. Most stations
aim for a small link margin, but the specific calculation of link margin and the determination of a
particular desired value are not specified by the standard.
To make informed changes to transmission power, stations may request a radio link measurement.
Stations send an Action frame requesting a transmission report. The Action frame sent in response
contains a TPC Report element with two descriptive statistics. First, it contains the transmit power of
the report frame. Based on the transmission power, the receiver of the report can estimate the path
loss of the radio link. Second, the report frame contains a value for the link margin, which informs
the receiver of the ratio between the received power and the minimum acceptable power. If the
minimum acceptable power were, say, -70 dBm and the signal was received at -60 dBm, the link
margin would be reported as 10 dB. If the link margin is "too high," transmit power can be reduced.
If the link margin is "too low," the transmit power can be increased. Like many other components of
the 802.11 standard, "too high" and "too low" are left to the discretion of the software.
The standard is designed to ensure that the maximum transmission power is not exceeded, but it
does not restrict power selection in any way. There is no requirement for advanced functionality. That
said, it is conceivable that advanced access points supporting transmit power control may attempt to
keep track of the power required to reach each associated station, so that close-in stations require
less power than far-away stations.
Dynamic Frequency Selection (DFS)
In addition to the requirement for transmit power control, European regulations require that stations
avoid interfering with 5 GHz radar systems, as well as spread the power load across all available
channels. Accomplishing this is the task of 802.11's Dynamic Frequency Selection (DFS) mechanism.
Basic operation of DFS
Dynamic frequency selection consists of a number of procedures to enable 802.11 devices to change
the radio channel based on measurements and regulatory requirements. It can affect the initial
association procedure and ongoing network operation.
When stations first associate to the network, the Association Request frame includes a Supported
Channels information element, which communicates the channels supported by the station. Access
points may reject the association based on the content of the information element, though such
behavior is not specified by the standard. One approach would be to reject stations that support "too
few" channels, on the theory that it limits the ability of the access point to switch operation on to
different channels because it must move to a channel supported by all associated stations.
Once used on an operational network, DFS will periodically test the channel for potential interference
from other radio systems, most notably 5 GHz European radar systems. Testing the channel is
accomplished by stopping transmissions on the network, measuring for potential interference, and, if
necessary, advertising that the channel will change.
Quieting the channel
To perform tests on the radio channel, quiet periods or quiet intervals are used. Quiet intervals are a
time when all stations in the BSS do not transmit, which is helpful when making measurements for
potential interference from radar systems. Quiet periods are scheduled by the inclusion of the Quiet
information element in Beacon and Probe Response frames, and describe when and how long all
stations should cease transmissions. Only the most recently transmitted Quiet information is
effective. When multiple Quiet information elements are transmitted, the most recent one supercedes
all prior scheduled quiet periods. During the quiet period, all stations set the network allocation
vector (NAV) to the length of the quiet period to ensure that the virtual carrier sensing algorithm will
defer transmissions.
When an upcoming quiet interval is scheduled, the radio channel still operates under normal rules for
access to the radio medium, with the additional rule that any frame exchange must complete before
the start of the quiet period. If a scheduled frame exchange cannot be completed, the station
relinquishes control of the channel and defers transmission until after the conclusion of the quiet
period. Failure to transmit a frame due to an impending quiet period does not, however, increase the
transmission count. When the quiet period resumes, all stations must contend for access to the radio
channel again. There is no preservation of channel access across a quiet period.
In an infrastructure network, channel quiet scheduling is completely under the control of the access
point. Access points are allowed to change the duration of quiet periods, the time between quiet
periods, or even to stop scheduling quiet periods altogether. Independent networks choose the quiet
period scheduling when the network is created. When new stations take over responsibility for
sending Beacon and Probe Response frames, they have no discretion to alter the quiet period
parameters, and simply copy the previously used parameters.
Measuring
At any point, measurements of the radio channel can be taken. Stations may request other stations
to measure the radio channel. Access points may find measurements from stations to be particularly
useful in learning about the state of the radio channel because the reports from stations are likely to
come from a variety of different geographical locations. Measurements may be taken in a quiet
period or while the radio is in service.
Any station may ask other stations to make a measurement. Inquiries for radio information are sent
in Measurement Request frames, described in Figure 8-23. In an infrastructure network, all frames
must go through the access point. Associated client stations may only ask the AP for radio
information. Access points in an infrastructure network may ask either a single station or a group of
stations for a measurement, simply by addressing the request frame appropriately. In independent
networks, there is no centralized point of control, and any station may issue a request to any other
single station or group of stations. Although requests to a group address field are allowed, receiving
stations are also allowed to disregard them.
After sending a measurement request, the standard assumes that the receiver needs time to gather
the data for its reply. Following the transmission of a measurement request, a station refrains from
sending any further frames.
Upon receipt of a Measurement Request frame, a station must determine how to respond. The
Measurement Request frame must always be answered, even if the response is a refusal to perform
the requested measurement. To be processed, a request must be received in enough time to set up
and take the measurement. Measurement requests specify a time at which the measurement is to
occur. If a request is held in a long transmission queue, it is conceivable that it could be arrive at the
destination after the requested measurement should begin. Stations are allowed to ignore such "late"
requests. The receiver of a measurement request must also be able to collect the data supported in
the request. Depending on the receiver's hardware, it may not be possible to support all requests.
Although increasingly rare, not all 802.11 hardware is capable of supporting every allowed
channel.[*] Many cards on the market today do not support multichannel operation, either, so a
request to measure a different channel from the current operating channel cannot be supported.
Stations may refuse to perform a measurement for any nonspecified reason, provided the requester
has not marked the request as mandatory.
[*]
For example, the Cisco CB-20 card, based on the Radiata chipset, could only support the eight lowest 802.11a channels. It
would be appropriate for such a card to refuse a measurement request in the high U-NII band.
In addition to the polled operation, in which stations ask for measurements, it is possible for stations
to spontaneously report statistics by sending unsolicited Measurement Report frames.
Radar scan
One of the major reasons to quiet the channel is to search for the presence of 5 GHz radar systems
in use in Europe. The regulations[*] do not require any particular search method for these radar
systems. It only requires that radar be detected when its signal strength rises above a defined
interference threshold.
[*]
ETSI EN 301 893, available from http://www.etsi.org.
When a radio interface is started, it must search for a radar signal on the channel. No transmissions
are allowed until the "coast is clear" and it has been established that there is no radar to interfere
with. Radar detection must be carried out periodically throughout operation. Whenever radar signals
are detected, the network must switch to another channel to avoid interference.
Spectrum management services enable a network to move to another channel. The decision to move
to another channel may be caused by the presence of radar interference, but the generic mechanism
to switch channels may be useful for a variety of purposes beyond compliance with European radio
regulations. Networks capable of altering their operational channel may do so to minimize
interference with other 802.11 devices and optimize the radio plan. Channel switching is designed to
move as many of the associated stations to the new channel as possible, but it is always possible
that communications with some or all of the associated stations may be disrupted. The standards do
not place any constraints on how to select the new channel, but simply say that an access point
should attempt to choose a channel supported by as many of its stations as possible. Some
regulators have drafted rules that mandate energy spreading across the entire band, so it may be
that regulatory requirements force a switch to channels that are not supported by all stations.
In an infrastructure network, the selection of the operating channel is under the sole control of the
access point. As part of the association process, access points collect information about which
channels can be supported by the associated stations. Access points inform associated stations of the
impending switch by using the Channel Switch Announcement information element in management
frames, as well as Action frames. To improve the ability of an access point to send an channel switch
announcement, the appropriate action frame may be sent after the PCF Interframe Space (PIFS),
which gives it a higher priority than any new atomic exchange on the medium. The standard
suggests, but does not require, that the channel switch be scheduled far enough in advance that any
powersaving stations have the opportunity to become active and receive the channel switch
announcement.
IBSS operation
Changing the operating channel in an independent network is significantly more complicated than in
an infrastructure network because there is no single logical controller for frequency selection
operation. Instead of having the function reside in the access point, independent networks have the
DFS owner service, which coordinates multiple stations to run the frequency selection service.
One station in a network is designated the DFS owner, and is responsible for collecting measurement
reports and monitoring the channel for radar signals. If any station in the independent network
observes a radar signal, it will be reported in the channel map subfield. Upon notification that radar
has been detected, the DFS owner takes charge of changing the channel.
The DFS owner is responsible for deciding on the new channel, and sending the channel switch
announcement frame. It may not be possible to select a channel that complies with regulatory
requirements and is supported by all stations. Independent networks do not have a central point of
data collection, so even if there is a channel supported by all stations, there is no guarantee that the
DFS owner will be aware of it.
In an independent network, the DFS owner may change. Just like Beacon generation, stations may
join or leave the network, including the DFS owner. To cope with the DFS owner going away, stations
may enter DFS owner recovery mode. In this mode, multiple stations can become DFS owners, and
will schedule the channel switch announcement frames required by the standard. However, the first
station to transmit the channel switch frame becomes the only DFS owner, and other stations will
drop the role. DFS owner recover is similar in concept to the distributed Beacon generation discussed
in Figure 8-20.
Action Frames
Action frames are used to request a station to take action on behalf of another. Spectrum
management services use Action frames to request that measurements be taken, gather the results
of those measurements, and announce any required channel switches. Figure 8-22 shows the format
of the Action frame, which is essentially a category plus details that depend on the category. The
action details may vary depending on the category field.
Figure 8-22. Action frame
Category
The category field is set to zero for spectrum management.
Action
All spectrum management frames use the first byte of the action details to specify the type of
action being undertaken. Table 8-2 shows the possible values for the Action subfield. Values
not shown are reserved.
Elements
Spectrum management action frames carry information in information elements. The details for
many information elements can be found in Chapter 4. Several additional information elements
were defined in 802.11h and are described in this section.
Table 8-2. Types of spectrum management action frames
Value Type of spectrum management action frame
0
Measurement Request
Value Type of spectrum management action frame
1
Measurement Report
2
TPC Request
3
TPC Report
4
Channel Switch Announcement
Measurement Request frame
The Measurement Request frame is used to request that a station make measurements and send the
results to the sender. Its format is shown in Figure 8-23. The frame consists of a series of
measurement request information elements. The number of potential measurements is limited by the
size of the frame, rather than any aspect of the frame construction.
Periodic measurements are allowed by the standard. Periodic reports are enabled or disabled by
sending a measurement request to a station with instructions to turn the periodic measurement on or
off. Stations cannot disable measurements on access points in infrastructure networks.
Category
Set to zero to indicate spectrum management action frames.
Action
Set to zero to indicate that it is a measurement request.
Dialog Token
This field acts like a sequence number. It is set to a non-zero value to assist in matching
measurement responses to outstanding requests.
A single Measurement Request frame may request multiple measurements by using multiple
Measurement Request information elements within the body of the frame.
Figure 8-23. Measurement Request frame
A single information element is shown in exploded view, consisting of the following fields:
Element ID
Measurement Request elements are type 38.
Length
The length, in bytes, of the information element following this field.
Measurement Token
Each Measurement Request frame may make several requests by including several
Measurement Request elements within the frame body. Each request is given its own value for
the Measurement Token so that the different requests can be distinguished.
Measurement Request Mode bitmap
There are three bits in the Measurement Request Mode bitmap that are used to indicate what
types of spectrum management frames are supported. Bit number 2 (numbered starting from
0) is the Request bit, and is set to one to signify that the transmitter will process incoming
measurement requests. Bit number 3 is the Report bit, which is set to one to signify that the
transmitter will accept unsolicited reports. The Enable bit is set to one when the other two bits
are valid.
Measurement Type
The type of measurement being requested by the information element, as shown in Table 8-3.
Measurement Request
If a measurement is requested, there may be an additional field to give timing parameters. As
it turns out, the three types of measurements that are currently standardized all have the
same format, consisting of a channel number, the value of the timer function at which the
measurement should start, and the duration of the measurement in time units. A timer start
value of zero indicates the measurement should be taken immediately. This field is not present
when the frame is used to turn measurements on or off.
Table 8-3. Measurement Type values
Measurement Type value
Name
0
Basic measurement
1
Clear channel assessment
2
Receive power indication (RPI) histogram
Measurement Report
The Measurement Report frame is used to send the results of a measurement to the requester. Its
format is shown in Figure 8-24. The frame consists of a series of measurement report information
elements. The number of potential reports is limited by the size of the frame, rather than any aspect
of the frame construction.
Category
Set to zero to indicate spectrum management action frames.
Action
Set to one to indicate that it is a measurement report.
Dialog Token
If the measurements were taken as the result of a measurement request from another station,
the Dialog Token field from that request is copied into the response. If the frame is sent as an
unsolicited report, the Dialog Token is zero.
A single Measurement Report frame may contain the results for several measurements, each
transmitted in its own information element. For clarity, a single information element is shown in the
information element header, and the three possible report elements are shown below.
Element ID
Measurement Request elements are type 39.
Length
The length, in bytes, of the information element following this field.
Measurement Token
Each Measurement Request frame may make several requests by including several
Measurement Request elements within the frame body. Each request is
Figure 8-24. Measurement Report frame
given its own value for the Measurement Token so that the different requests can be
distinguished.
Measurement Report Mode bitmap
There are three bits in the Measurement Report Mode bitmap that are used to indicate why a
measurement is being refused, if the report frame is being used to deny a measurement. The
Late bit is set to one when the measurement request arrives after the start time specified in
the measurement. The Incapable bit is set to one when a station is unable to perform the
requested measurement. The Refused bit is set to one when the station is capable of
performing the measurement, but does not wish to do so.
Measurement Type
The type of measurement being requested by the information element, as shown in Table 8-3.
Measurement Report
The Measurement Report frame contains the requested measurement data. Unlike the
Measurement Request, the contents are different for each type of report. All three reports are
shown underneath the information element header. All share a common header that reports
the channel number that the request is for, the time the measurement started, and the
duration of the measurement. However, the data is reported in different ways for each type of
measurement.
In a basic report, the data reported are a series of bit flags for the channel:
BSS (1 bit)
This bit will be set if frames from another network are detected during a measurement period.
OFDM Preamble (1 bit)
This bit is set if the 802.11a short training sequence is detected, but without being followed by
the rest of the frame. HIPERLAN/2 networks use the same preamble, but not the same frame
construction.
Unidentified Signal (1 bit)
This bit is set when the received power is high, but the signal cannot be classified as either
another 802.11 network (and hence, set the BSS bit), another OFDM network (and hence, set
the OFDM Preamble bit), or a radar signal (and hence, set the Radar bit). The standard does
not specify what power level is high enough to trigger this bit being set.
Radar (1 bit)
If a radar signal is detected during a measurement period, this bit will be set. Radar systems
that must be detected are defined by regulators, not the 802.11 task group.
Unmeasured (1 bit)
If the channel was not measured, this bit will be set. Naturally, if there was no measurement
taken, nothing can be detected in the band and the previous four bits will be set to zero.
In a CCA report, the main field is the CCA Busy Fraction, which describes the fraction of time on
which the clear channel assessment function was set to busy. It is a single byte, so the fraction is
multiplied by 255 to convert to an integer ranging from 0 to 255, where higher values indicate the
channel was busy more often.
RPI histogram reports are used to report the spread of received power on an interface. Stations can
request an RPI histogram to determine how well another station is able to see signals from the
current network, or it may use the report to scout out other channels when it is time to change the
operating channel. An RPI histogram report contains information on the strength of received signals.
However, unlike a one-frame measurement, it can report the spread of power received over the
measurement duration to give the receiver an indication of the overall level of transmissions. The
histogram contains eight bytes, each of which represents a range of received power, as shown in
Table 8-4. Each byte has a value representing the fraction of power signals that fall into its range.
The time fraction over which the received signal falls into the power range for the byte is scaled so
that each byte ranges from 0 to 255, with its value depending on the fraction of time signals were
received at that power level.[*]
[*]
Although the values of all 8 bytes theoretically add to 255, the standard notes that they may add up to 262 due to rounding
effects.
Table 8-4. RPI power mapping
RPI number
Corresponding power (dBm)
0
Less than -87
1
-87 to -82
2
-81 to -77
3
-78 to -72
4
-71 to -67
5
-66 to -62
6
-61 to -57
7
-56 and higher
TPC Request and Report
TPC Request and TPC Report frames are both shown in Figure 8-25. They are both straightforward,
consisting of Action frames with the spectrum management type. Each frame contains its
corresponding information element, as described in Chapter 4. As with other frames, the Dialog
Token field is used to match up requests with responses.
Figure 8-25. TPC Request and Report frames
Channel Switch Announcement
When the channel must be changed, stations that are part of the network must be informed of the
impending change so that they may prepare to switch to the specified new channel. Channel Switch
Announcement frames, shown in Figure 8-26, are essentially Action frame wrappers around the
Channel Switch Announcement element that was described in Chapter 4. As such, they have all the
functions of a channel switch announcement element, and are used to specify the time at which the
network will switch to a new channel.
Figure 8-26. Channel Switch Announcement frame
Chapter 9. Contention-Free Service with the
PCF
To support applications that require near real-time service, the 802.11 standard includes a second
coordination function to provide a different way of accessing the wireless medium. The point
coordination function (PCF) allows an 802.11 network to provide an enforced "fair" access to the
medium. In some ways, access to the medium under the PCF resembles token-based medium access
control schemes, with the access point holding the token. This chapter describes medium access
under the PCF, detailed frame diagrams for the PCF frames, and how power management operations
interact with the PCF.
The PCF has not been widely implemented. A media server product for the home implemented the
PCF, though it was not commercially successful. Some enterprise-class products have implemented
the PCF because it gives the access points more control over access to the wireless medium, and
helps the network to to wrest control away from the anarchy of a herd of individual stations. For
many readers, this chapter may not be necessary. If you are not using a product that implements the
PCF, there is no need to read this chapter unless you have an interest in the standard itself.
Contention-Free Access Using the PCF
If contention-free delivery is required, the PCF may be used. The PCF is an optional part of the
802.11 specification; products are not required to implement it. However, the IEEE designed the PCF
so stations that implement only the distributed coordination function (DCF) will interoperate with
point coordinators.
Contention-free service is not provided full-time. Periods of contention-free service arbitrated by the
point coordinator alternate with the standard DCF-based service. The relative size of the contentionfree period can be configured. 802.11 describes the contention-free periods as providing "near
isochronous" services because the contention-free periods will not always start at the expected time,
as described in the "Contention-Free Period Duration" section later in this chapter.
Contention-free service uses a centralized access control method. Access to the medium is restricted
by the point coordinator, a specialized function implemented in access points. Associated stations can
transmit data only when they are allowed to do so by the point coordinator. In some ways,
contention-free access under the PCF resembles token-based networking protocols, with the point
coordinator's polling taking the place of a token. Fundamentals of the 802.11 model remain in place,
however. Although access is under the control of a central entity, all transmissions must be
acknowledged.
PCF Operation
Figure 9-1 shows a transfer using the PCF. When the PCF is used, time on the medium is divided into
the contention-free period (CFP) and the contention period. Access to the medium in the former case
is controlled by the PCF, while access to the medium in the latter case is controlled by the DCF and
the rules from Chapter 8. The contention period must be long enough for the transfer of at least one
maximum-size frame and its associated acknowledgment. Alternating periods of contention-free
service and contention-based service repeat at regular intervals, which are called the contention-free
repetition interval.
Figure 9-1. Using the PCF
Reserving the medium during the contention-free period
At the beginning of the contention-free period, the access point transmits a Beacon frame. One
component of the beacon announcement is the maximum duration of the contention-free period,
CFPMaxDuration. All stations receiving the Beacon set the NAV to the maximum duration to lock out
DCF-based access to the wireless medium.
As an additional safeguard to prevent interference, all contention-free transmissions are separated
only by the short interframe space and the PCF interframe space. Both are shorter than the DCF
interframe space, so no DCF-based stations can gain access to the medium using the DCF.
The polling list
After the access point has gained control of the wireless medium, it polls any associated stations on a
polling list for data transmissions. During the contention-free period, stations may transmit only if
the access point solicits the transmission with a polling frame. Contention-free polling frames are
often abbreviated CF-Poll. Each CF-Poll is a license to transmit one frame. Multiple frames can be
transmitted only if the access point sends multiple poll requests.
The polling list is the list of privileged stations solicited for frames during the contention-free period.
Stations get on the polling list when they associate with the access point. The Association Request
includes a field that indicates whether the station is capable of responding to polls during the
contention-free period.
Transmissions from the Access Point
Generally, all transmissions during the contention-free period are separated by only the short
interframe space. To ensure that the point coordinator retains control of the medium, it may send to
the next station on its polling list if no response is received after an elapsed PCF interframe space.
(Such a situation is illustrated in Figure 9-1.) The access point polls the second station on its list but
receives no response. After waiting one PCF interframe space, the access point moves to the third
station on the list. By using the PCF interframe space, the access point ensures that it retains access
to the medium.
The access point may use several different types of frames during the contention-free period. During
this period, the point coordinator has four major tasks. In addition to the "normal" tasks of sending
buffered frames and acknowledging frames from the stations, the point coordinator can poll stations
on the polling list to enable them to send frames; it may also need to transmit management frames.
Time in the contention-free period is precious, so acknowledgments, polling, and data transfer may
be combined to improve efficiency. When any subset of these functions are combined into a single
frame, the result is a bit strange. A single frame could, for example, acknowledge the receipt of the
previous frame, poll a different station for buffered data, and send its own data to the station on the
polling list.
Several different frame types can be used in the contention-free period:
Data
The standard vanilla Data frame is used when the access point is sending a frame to a station
and does not need to acknowledge a previous transmission. The standard Data frame does not
poll the recipient and thus does not allow the recipient to transmit any data in return. The data
only frame used in the contention-free period is identical to the Data frame used in contentionbased periods.
CF-Ack
This frame is used by stations to acknowledge the receipt of a frame when no data needs to be
transmitted. Contention-free acknowledgments are longer than the standard control frame
acknowledgment, so this frame may not be used in actual implementations.
CF-Poll
CF-Poll frames are sent by the access point to a mobile station to give the mobile station the
right to transmit a single buffered frame. It is used when the access point does not have any
data for the mobile station. When a frame for the mobile station is available, the access point
uses the Data+CF-Poll frame type.
Data+CF-Ack
This frame combines data transmission with an acknowledgment. Data is directed to the frame
recipient; the acknowledgment is for the previous frame transmitted and usually is not for the
recipient of the data.
Data+CF-Poll
This frame is used by access points to transmit data to a mobile station and request one
pending frame from the mobile station. The Data+CF-Poll can only be sent by the access point
during the contention-free period.
CF-ACK+CF-Poll
This frame acknowledges the last frame from one of the access point's clients and requests a
buffered frame from the next station on the polling list. It is directed to the next station on the
polling list, although the acknowledgment may be intended for any mobile station associated
with the access point.
Data+CF-ACK+CF-Poll
This frame brings together the data transmission, polling feature, and acknowledgment into
one frame for maximum efficiency.
CF-End
This frame ends the contention-free period and returns control of the medium to the
contention-based mechanisms of the DCF.
CF-End+CF-Ack
This is the same as the CF-End frame but also acknowledges the previously transmitted Data
frame.
Any Management
No restriction is placed by the standard on which management frames can be transmitted
during the contention-free period. If the rules applying to a particular frame type allow its
transmission, the access point may transmit it.
Contention-Free Period Duration
The minimum length of the contention period is the time required to transmit and acknowledge one
maximum-size frame. It is possible for contention-based service to overrun the end of the contention
period, however. When contention-based service runs past the expected beginning of the contentionfree period, the contention-free period is foreshortened, as in Figure 9-2.
Figure 9-2. Data+CF-Ack and Data+CF-Poll usage
When the contention-free period is foreshortened, the existing frame exchange is allowed to
complete before the beacon announcing the start of contention-free operation is transmitted. The
contention-free period is shortened by the amount of the delay. Contention-free service ends no later
than the maximum duration from the expected beginning point, which is referred to as the Target
Beacon Transmission Time (TBTT).
The point coordinator may also terminate the contention-free period prior to its maximum duration
by transmitting a CF-End frame. It can base this decision on the size of the polling list, the traffic
load, or any other factor that the access point considers important. Products will sometimes
selectively use the PCF for a particular frame exchange by starting contention-free service,
performing the desired frame transmission, and then ending it.
Detailed PCF Framing
Several frame types are used exclusively within the contention-free period. They combine, in various
states, data transmission, acknowledgment, and polling. This section describes when various frames
are used and how the different functions interact during frame exchanges. Contention-free frames
combine several functions into a single frame.
Data+CF-Ack
The Data+CF-Ack frame combines two different functions for transmission efficiency. Data is
transmitted in the frame payload, and the frame implicitly acknowledges the receipt of data
received one short interframe space previously. Generally, the data and the acknowledgment
are intended for two separate stations. In Figure 9-3, the contention-free acknowledgment is
coupled with the data for transmission to the access point in the previous frame, but the data
may be intended for any station on the 802.11 network.
Figure 9-3. Data+CF-Ack usage
This frame is used only in infrastructure networks because it is transmitted during the
contention-free period. It may be transmitted by either the access point or a mobile station.
During the contention-free period, however, the access point is responsible for polling, and it is
unlikely that it would transmit this frame subtype because it does not include a poll.
Data+CF-Poll
The Data+CF-Poll frame is used by access points in infrastructure networks during the
contention-free period. When the access point does not need to acknowledge any outstanding
frames, it sends a Data+CF-Poll to transmit data to the recipient and allows the recipient to
send one buffered frame in response. The data in the frame body must be intended for the
recipient of the poll; the two operations cannot be "split" across two different receivers. In
Figure 9-3, the access point uses a Data+CF-Poll frame to send one frame to the mobile station
and to solicit the response.
Data+CF-Ack+CF-Poll
The Data+CF-Ack+CF-Poll frame is used by access points in infrastructure networks during the
contention-free period. When the access point has data to transmit, must acknowledge a
frame, and needs to poll a station on the polling list, all the functions can be combined into one
frame. Figure 9-4 illustrates the usage of Data+CF-Ack+CF-Poll. As with Data+CF-Ack, the
components of the Data+CF-Ack+CF-Poll frame are generally intended for different stations.
The data transmission and polling must be intended for the same station, but the
acknowledgment is for the previous transmission.
Figure 9-4. Usage of Data+CF-Ack+CF-Poll
The figure begins with mobile station 1 (MS1) transmitting a Data+CF-Ack frame. The Data
must go to the access point, but the CF-Ack is used to acknowledge the previous Data frame
transmitted by the access point. (That frame is not shown in the figure.) Moving down the
polling list, the access point then polls mobile station 2 (MS2). However, the access point must
acknowledge the data from MS1, which it does by transmitting a frame with a CF-Ack
component. When the access point also has data to transmit, all three features can be
combined into one omnibus frame. The Data and CF-Poll components are intended for the
recipient of the frame, but the CF-Ack is intended for the transmitter of the previous frame.
MS1 must listen to the access point frames to note the acknowledgment.
CF-Ack (no data)
When only an acknowledgment is required, a header-only frame with just the CF-Ack function
can be transmitted. In Figure 9-4, if MS2 had no data to transmit, it would have responded
with a CF-Ack frame.
CF-Poll (no data)
CF-Poll can also be transmitted by itself. Naturally, only access points perform this function, so
the CF-Poll frame is transmitted only by access points in infrastructure networks during the
contention-free period.
"Naked" CF-Polls are transmitted when the access point has no buffered data for the recipient
and does not need to acknowledge the receipt of previous frames. One common situation in
which no acknowledgment is necessary is when the access point transmits a CF-Poll and the
polled station has no data and does not respond. If the access point has no data for the next
station on the polling list, it transmits a CF-Poll, as in Figure 9-5.
Figure 9-5. CF-Poll framing usage
In Figure 9-5, the access point attempts to transmit data to MS1 but does not receive a
response. After the PCF interframe space has elapsed, the access point can proceed down the
polling list to MS2. No frame from MS1 needs to be acknowledged, and if the access point has
no data for MS2, it can use a CF-Poll to allow MS2 to send data.
CF-Ack+CF-Poll (no data)
The final subtype of Data frame is the CF-Ack+CF-Poll, which is also transmitted by access
points. Like all CF-Poll frames, it is used only during the contention-free period and only by
access points. It incorporates the acknowledgment function and the polling function into a
frame with no data. Figure 9-6 illustrates its usage.
Figure 9-6. CF-Ack+CF-Poll usage
The scenario is a slight variation on the previous setting. Instead of a timeout waiting for MS1
to respond, MS1 returns a frame. When the access point takes control of the medium, it uses a
CF-Ack+CF-Poll to acknowledge receipt of the frame from MS1 and notifies MS2 that it is
allowed to send a frame.
Contention-Free End (CF-End)
When the contention-free period ends, the access point transmits a CF-End frame to release stations
from the PCF access rules and begin contention-based service. The format of the CF-End frame is
shown in Figure 9-7. Four fields make up the MAC header of the CF-End frame:
Frame Control
The frame subtype is set to 1110 to indicate a CF-End frame.
Duration
CF-End announces the end of the contention-free period and thus does not need to extend the
virtual carrier sense. Duration is set to 0. Stations that receive the CF-End frame cut the virtual
carrier sense short to resume contention-based access.
Address 1: Receiver Address
CF-End is relevant to the operation of all mobile stations, so the receiver address is the
broadcast address.
Address 2: BSSID
CF-End is announced by the access point to all the stations associated with its BSS, so the
second address field is the BSSID. In infrastructure networks, the BSSID is the address of the
wireless interface in the access point, so the BSSID is also the transmitter address.
Figure 9-7. CF-End frame
CF-End+CF-Ack
When the contention-free period ends, the access point transmits a CF-End frame to release stations
from the PCF access rules and then begins contention-based service using the DCF. If the access
point must also acknowledge receipt of data, it may simultaneously end the contention-free period
and acknowledge the previous frame by using the CF-End+CF-Ack frame, which combines both
functions. The format of the CF-End+CF-Ack frame is shown in Figure 9-8. Four fields make up the
MAC header of the CF-End+CF-Ack frame:
Frame Control
The frame subtype is set to 1111 to indicate a CF-End+CF-Ack frame.
Duration
CF-End+CF-Ack announces the end of the contention-free period and thus does not need to
extend the virtual carrier sense. Duration is set to 0.
Address 1: Receiver Address
CF-End+CF-Ack is relevant to the operation of all mobile stations, so the receiver address is
the broadcast address.
Address 2: BSSID
CF-End+CF-Ack is announced by the access point to all the stations associated with its BSS, so
the second address field is the BSSID. In infrastructure networks, the BSSID is the address of
the wireless interface in the access point, so the BSSID is also the transmitter address.
Figure 9-8. CF-End+CF-Ack frame
CF Parameter Set
Access points that support contention-free operation may include the CF Parameter Set information
element, which is shown in Figure 9-9. CF Parameter Set elements are included in Beacon frames to
keep all mobile stations apprised of contention-free operations. They are also included in Probe
Response frames to allow stations to learn about contention-free options supported by a BSS. Four
fields make up the CF Parameter Set information element:
CFP Count
This field, which is one byte in length, tells how many DTIM frames will be transmitted before
the start of the next contention-free period. Zero indicates that the current frame is the start of
contention-free service.
CFP Period
This one-byte field indicates the number of DTIM intervals between the start of contention-free
periods.
CFP MaxDuration
This value is the maximum duration of the contention-free period as measured in time units
(TUs). Mobile stations use this value to set the NAV to busy for the entire contention-free
period.
CFP DurRemaining
This value is the number of TUs remaining in the current contention-free period. Mobile
stations use it to update the NAV throughout the contention-free period. When DCF-based
contention-free service is provided, it is set to 0.
Figure 9-9. CF Parameter Set information element
Power Management and the PCF
Power conservation during the contention-free period is similar to power conservation during the
contention-based period, with a few minor exceptions. The basic distinction between the two is that
frame delivery must obey the PCF rules, so buffered frames can be delivered only to CF-Pollable
stations. Stations that do not support PCF operations must wait until contention-based service
resumes before retrieving buffered frames.
Stations on the polling list are not allowed to sleep during the contention-free period. When the
access point is performing its point coordination functions, it may poll any station on the polling list
at any time. Frames destined for stations on the polling list do not need to be buffered during the
contention-free period because those stations do not sleep.
Frame buffering is identical under contention-free and contention-based service. By maintaining
power-saving status for each station, the access point can buffer frames for any station in a lowpower mode. Broadcast and multicast frames are buffered whenever an associated station is in a
low-power mode.
In addition to the buffer status associated with contention-free service, the access point also sets bits
in the TIM for any station it intends to poll. The reason for setting these bits is related to how
buffered frames are delivered. Like contention-based service, DTIM frames trigger the transmission
of broadcast and multicast frames. If the total time required to transmit multicast and broadcast
frames exceeds the Beacon interval, the access point will transmit one Beacon interval's worth of
buffered frames and stop. Remaining frames will, however, cause the access point to keep the bit
corresponding to AID 0 set.
After transmitting the buffered broadcast and multicast frames, the access point goes through the list
of AIDs whose TIM bits are set in increasing order and transmits any pending data. Transmissions
are conducted according to the rules of the PCF, so it is not necessary to include a delay before
beginning transmission. Stations on the polling list are added to the TIM, so they will be included in
this process. Multiple buffered frames can be transmitted, but this is entirely up to the access point
implementationin contention-free service, mobile stations can transmit only when given permission
by the access point. A station is not allowed to resume sleeping until all frames have been delivered
to it, as indicated by a 0 More Data bit. When a station is cleared to resume sleeping, it sleeps until
the next DTIM transmission. DTIM frames signal the beginning of the contention-free period, so all
stations that implement the PCF are required to wake up for every DTIM.
If a station switches from a low-power mode to the active mode, any frames buffered for it are
transferred to the point coordination function for delivery during the contention-free period. The
transfer does not result in immediate delivery, but the access point can place the frames into a
queue for transmission as soon as the point coordination function permits.
Chapter 10. Physical Layer Overview
Any girl can be glamorous.
All you have to do is stand still and look stupid.
Hedy Lamarr
Protocol layering allows for research, experimentation, and improvement on different parts of the
protocol stack. The second major component of the 802.11 architecture is the physical layer, which is
often abbreviated PHY. This chapter introduces the common themes and techniques that appear in
each of the radio-based physical layers and describes the problems common to all radio-based
physical layers; it is followed by more detailed explanations of each of the physical layers that are
standardized for 802.11.
Physical-Layer Architecture
The physical layer is divided into two sublayers: the Physical Layer Convergence Procedure (PLCP)
sublayer and the Physical Medium Dependent (PMD) sublayer. The PLCP (Figure 10-1) is the glue
between the frames of the MAC and the radio transmissions in the air. It adds its own header.
Normally, frames include a preamble to help synchronize incoming transmissions. The requirements
of the preamble may depend on the modulation method, however, so the PLCP adds its own header
to any transmitted frames. The PMD is responsible for transmitting any bits it receives from the PLCP
into the air using the antenna. The physical layer also incorporates a clear channel assessment (CCA)
function to indicate to the MAC when a signal is detected.
Figure 10-1. Physical layer logical architecture
The Radio Link
Three physical layers were standardized in the initial revision of 802.11, which was published in
1997:
Frequency-hopping (FH) spread-spectrum radio PHY
Direct-sequence (DS) spread-spectrum radio PHY
Infrared light (IR) PHY
Later, three further physical layers based on radio technology were developed:
802.11a: Orthogonal Frequency Division Multiplexing (OFDM) PHY
802.11b: High-Rate Direct Sequence (HR/DS or HR/DSSS) PHY
802.11g: Extended Rate PHY (ERP)
The future 802.11n, which is colloquially called the MIMO PHY or the High-Throughput PHY
This book discusses the physical layers based on radio waves in detail; it does not discuss the
infrared physical layer, which to my knowledge has never been implemented in a commercial
product.
The IR PHY
It is often said that microwave ovens operate at 2.45 GHz because it corresponds to a
particular excitation mode of water molecules. This is sometimes even offered as a
reason why 802. 11 cannot be used over long distances. If atmospheric water vapor
would severely attenuate any microwave signals in rain or in humid climates, then
802.11 is not suitable for use over long distances.
802.11 also includes a specification for a physical layer based on infrared (IR) light.
Using infrared light instead of radio waves seems to have several advantages. IR ports
are less expensive than radio transceiversin fact, the cost is low enough that IR ports are
standard on practically every laptop.
IR is extremely tolerant of radio frequency (RF) interference because radio waves
operate at a totally different frequency. This leads to a second advantage: IR is
unregulated. Product developers do not need to investigate and comply with directives
from several regulatory organizations throughout the world.
Security concerns regarding 802.11 are largely based on the threat of unauthorized
users connecting to a network. Light can be confined to a conference room or office by
simply closing the door. IR-based LANs can offer some of the advantages of flexibility
and mobility but with fewer security concerns. This comes at a price. IR LANs rely on
scattering light off the ceiling, so range is much shorter.
This discussion is academic, however. No products have been created based on the IR
PHY. The infrared ports on laptops comply with a set of standards developed by the
Infrared Data Association (IrDA), not 802.11. Even if products were created around the
IR PHY, the big drivers to adopt 802.11 are flexibility and mobility, which are better
achieved by radio's longer range and ability to penetrate opaque objects.
Licensing and Regulation
The classic approach to radio communications is to confine an information-carrying signal to a narrow
frequency band and pump as much power as possible (or legally allowed) into the signal. Noise is
simply the naturally present distortion in the frequency band. Transmitting a signal in the face of
noise relies on brute forceyou simply ensure that the power of the transmitted signal is much greater
than the noise.
In the classic transmission model, avoiding interference is a matter of law, not physics. With high
power output in narrow bands, a legal authority must impose rules on how the RF spectrum is used.
In the United States, the Federal Communications Commission (FCC) is responsible for regulating the
use of the RF spectrum. Many FCC rules are adopted by other countries throughout the Americas.
European allocation is performed by the European Radiocommunications Office (ERO) and the
European Telecommunications Standards Institute (ETSI). In Japan, the Ministry of Internal
Communications (MIC) regulates radio usage. Worldwide "harmonization" work is often done under
the auspices of the International Telecommunications Union (ITU). Many national regulators will
adopt ITU recommendations.
For the most part, an institution must have a license to transmit at a given frequency. Licenses can
restrict the frequencies and transmission power used, as well as the area over which radio signals
can be transmitted. For example, radio broadcast stations must have a license from the FCC.
Likewise, mobile telephone networks must obtain licenses to use the radio spectrum in a given
market. Licensing guarantees the exclusive use of a particular set of frequencies. When licensed
signals are interfered with, the license holder can demand that a regulatory authority step in and
resolve the problem, usually by shutting down the source of interference. Intentional interference is
equivalent to trespassing, and may be subject to criminal or civil penalties.
Frequency allocation and unlicensed frequency bands
Radio spectrum is allocated in bands dedicated to a particular purpose. A band defines the
frequencies that a particular application may use. It often includes guard bands, which are unused
portions of the overall allocation that prevent extraneous leakage from the licensed transmission
from affecting another allocated band.
Several bands have been reserved for unlicensed use. For example, microwave ovens operate at
2.45 GHz, but there is little sense in requiring homeowners to obtain permission from the FCC to
operate microwave ovens in the home. To allow consumer markets to develop around devices built
for home use, the FCC (and its counterparts in other countries) designated certain bands for the use
of "industrial, scientific, and medical" equipment. These frequency bands are commonly referred to
as the ISM bands. The 2.4-GHz band is available worldwide for unlicensed use.[*] Unlicensed use,
however, is not the same as unlicensed sale. Building, manufacturing, and designing 802.11
equipment does require a license; every 802.11 card legally sold in the U.S. carries an FCC
identification number. The licensing process requires the manufacturer to file a fair amount of
information with the FCC. Much this information is a matter of public record and can be looked up
online by using the FCC identification number.
[*]
The 2.4-GHz ISM band is reserved by the FCC rules (Title 47 of the Code of Federal Regulations), part 15.247. ETSI reserved
the same spectrum in ETSI Technical Specifications (ETS) 300-328.
The Nonexistent Microwave Absorption Peak of Water
Spread spectrum was patented in the early 1940s by Austrian-born actress Hedy Lamarr.
She was certainly better known for other reasons: appearing in the first nude scene on
film in the Czech film Ecstasy, her later billing as "the most beautiful woman in the
world" by Hollywood magnate Louis Mayer, and as the model for Catwoman in the
Batman comics.
It is often said that microwave ovens operate at 2.45 GHz because it corresponds to a
particular excitation mode of water molecules. This is sometimes even offered as a
reason why 802. 11 cannot be used over long distances. If atmospheric water vapor
would severely attenuate any microwave signals in rain or in humid climates, then
802.11 is not suitable for use over long distances.
The existence of a water excitation mode in the microwave range is a myth. If there was
an excitation mode, water would absorb a significant amount of the microwave energy.
And if that energy was absorbed effectively by water, microwave ovens would be unable
to heat anything other than the water near the surface of food, which would absorb all
the energy, leaving the center cold and raw. An absorption peak would also mean that
atmospheric water vapor would disrupt satellite communications, which is not an
observed phenomenon. NASA Reference Publication 1108(02), Propagation Effects on
Satellite Systems at Frequencies Below 10 GHz, discusses the expected signal loss due to
atmospheric effects, and the loss is much more pronounced at frequencies above 10
GHz. The absorption peak for water, for example, is at 22.2 GHz.
Microwave ovens do not work by moving water molecules into an excited state. Instead,
they exploit the unusually strong dipole moment of water. Although electrically neutral,
the dipole moment allows a water molecule to behave as if it were composed of small
positive and negative charges at either end of a rod. In the cavity of a microwave oven,
the changing electric and magnetic fields twist the water molecules back and forth.
Twisting excites the water molecules by adding kinetic energy to the entire molecule but
does not change the excitation state of the molecule or any of its components.
Use of equipment in the ISM bands is generally license-free, provided that devices operating in them
do not emit significant amounts of radiation. Microwave ovens are high-powered devices, but they
have extensive shielding to restrict radio emissions. Unlicensed bands have seen a great deal of
activity in the past three years as new communications technologies have been developed to exploit
the unlicensed band. Users can deploy new devices that operate in the ISM bands without going
through any licensing procedure, and manufacturers do not need to be familiar with the licensing
procedures and requirements. At the time this book was written, a number of new communications
systems were being developed for the 2.4-GHz ISM band:
The variants of 802.11 that operate in the band (the frequency-hopping layer, both direct
sequence layers, and the OFDM layer)
Bluetooth, a short-range wireless communications protocol developed by an industry
consortium led by Ericsson
Spread-spectrum cordless phones introduced by several cordless phone manufacturers
X10, a protocol used in home automation equipment that can use the ISM band for video
transmission
Unfortunately, "unlicensed" does not necessarily mean "plays well with others." All that unlicensed
devices must do is obey limitations on transmitted power. No regulations specify coding or
modulation, so it is not difficult for different vendors to use the spectrum in incompatible ways. As a
user, the only way to resolve this problem is to stop using one of the devices; because the devices
are unlicensed, regulatory authorities will not step in.
Other unlicensed bands
Additional spectrum is available in the 5 GHz range. The United States was the first country to allow
unlicensed device use in the 5 GHz range, though both Japan and Europe followed.[*] There is a large
swath of spectrum available in various countries around the world:
[*]
Europe is obviously not a single country, but there is a European-wide spectrum regulator.
4.92-4.98 GHz (Japan)
5.04-5.08 GHz (Japan)
5.15-5.25 GHz (United States, Japan)
5.25-5.35 GHz (United States)
5.47-5.725 GHz (United States, Europe)
5.725-5.825 GHz (United States)
Devices operating in 5 GHz range must obey limitations on channel width and radiated power, but no
further constraints are imposed. Japanese regulations specify narrower channels than either the U.S.
or Europe.
Spread Spectrum
Spread-spectrum technology is the foundation used to reclaim the ISM bands for data use.
Traditional radio communications focus on cramming as much signal as possible into as narrow a
band as possible. Spread spectrum works by using mathematical functions to diffuse signal power
over a large range of frequencies. When the receiver performs the inverse operation, the smearedout signal is reconstituted as a narrow-band signal, and, more importantly, any narrow-band noise is
smeared out so the signal shines through clearly.
Use of spread-spectrum technologies is a requirement for unlicensed devices. In some cases, it is a
requirement imposed by the regulatory authorities; in other cases, it is the only practical way to
meet regulatory requirements. As an example, the FCC requires that devices in the ISM band use
spread-spectrum transmission and impose acceptable ranges on several parameters.
Spreading the transmission over a wide band makes transmissions look like noise to a traditional
narrowband receiver. Some vendors of spread-spectrum devices claim that the spreading adds
security because narrowband receivers cannot be used to pick up the full signal. Any standardized
spread-spectrum receiver can easily be used, though, so additional security measures are mandatory
in nearly all environments.
This does not mean that spread spectrum is a "magic bullet" that eliminates interference problems.
Spread-spectrum devices can interfere with other communications systems, as well as with each
other; and traditional narrow-spectrum RF devices can interfere with spread spectrum. Although
spread spectrum does a better job of dealing with interference within other modulation techniques, it
doesn't make the problem go away. As more RF devices (spread-spectrum or otherwise) occupy the
area that your wireless network covers, you'll see the noise level go up, the signal-to-noise ratio
decrease, and the range over which you can reliably communicate drop.
To minimize interference between unlicenced devices, the FCC imposes limitations on the power of
spread-spectrum transmissions. The legal limits are one watt of transmitter output power and four
watts of effective radiated power (ERP). Four watts of ERP are equivalent to 1 watt with an antenna
system that has 6-dB gain, or 500 milliwatts with an antenna of 10-dB gain, etc.[*] The transmitters
and antennas in PC Cards are obviously well within those limitsand you're not getting close even if
you use a commercial antenna. But it is possible to cover larger areas by using an external amplifier
and a higher-gain antenna. There's no fundamental problem with doing this, but you must make sure
that you stay within the FCC's power regulations.
[*]
Remember that the transmission line is part of the antenna system, and the system gain includes transmission line losses. So
an antenna with 7.5-dB gain and a transmission line with 1.5-dB loss has an overall system gain of 6 dB. It's worth noting that
transmission line losses at UHF freqencies are often very high; as a result, you should keep your amplifier as close to the antenna
as possible.
The Unlikely Invention of Spread Spectrum
Spread spectrum was patented in the early 1940s by Austrian-born actress Hedy Lamarr.
She was certainly better known for other reasons: appearing in the first nude scene on
film in the Czech film Ecstasy, her later billing as "the most beautiful woman in the
world" by Hollywood magnate Louis Mayer, and as the model for Catwoman in the
Batman comics.
Before fleeing the advance of Nazi Germany, she was married to an Austrian arms
merchant. While occupying the only socially acceptable role available to her as a hostess
and entertainer of her husband's business clients, she learned that radio remote control
of torpedoes was a major area of research for armaments vendors. Unfortunately,
narrowband radio communications were subject to jamming, which neutralized the
advantage of radio-guided weapons. From these discussions, she first hit on the idea of
using a complex but predetermined hopping pattern to move the frequency of the control
signal around. Even if short bursts on a single frequency could be jammed, they would
move around quickly enough to prevent total blockage. Lamarr worked out everything
except how to precisely control the frequency hops.
After arriving in the United States, she met George Antheil, an avant-garde American
composer known as the "bad boy of music" for his dissonant style. His famous Ballet
mécanique used (among many outrageous noisemakers) 16 player pianos controlled
from a single location. Performing the piece required precisely controlled timing between
distributed elements, which was Lamarr's only remaining challenge in controlling the
hopping pattern. Together, they were granted U.S. patent number 2,292,387 in 1942.
The patent expired in 1959 without earning a cent for either of them, and Lamarr's
contributions went unacknowledged for many years because the name on the patent was
Hedy Kiesler Markey, her married name at the time. The emerging wireless LAN market
in the late 1990s led to the rediscovery of her invention and widespread recognition for
the pioneering work that laid the foundation for modern telecommunications.
Frequency-hopping techniques were first used by U.S. ships blockading Cuba during the
Cuban Missile Crisis. It took many years for the electronics underpinning spreadspectrum technology to become commercially viable. Now that they have, spreadspectrum technologies are used in cordless and mobile phones, high-bandwidth wireless
LAN equipment, and every device that operates in the unlicensed ISM bands.
Unfortunately, Hedy Lamarr died in early 2000, just as the wireless LAN market was
gaining mainstream attention.
Types of spread spectrum
The radio-based physical layers in 802.11 use three different spread-spectrum techniques:
Frequency hopping (FH or FHSS)
Frequency-hopping systems jump from one frequency to another in a random pattern,
transmitting a short burst at each subchannel. The 2-Mbps FH PHY is specified in clause 14.
Direct sequence (DS or DSSS)
Direct-sequence systems spread the power out over a wider frequency band using
mathematical coding functions. Two direct-sequence layers were specified. The initial
specification in clause 15 standardized a 2-Mbps PHY, and 802.11b added clause 18 for the
HR/DSSS PHY.
Orthogonal Frequency Division Multiplexing (OFDM)
OFDM divides an available channel into several subchannels and encodes a portion of the signal
across each subchannel in parallel. The technique is similar to the Discrete Multi-Tone (DMT)
technique used by some DSL modems. Clause 17, added with 802.11a, specifies the OFDM
PHY. Clause 18, added in 802.11g, specifies the ERP PHY, which is essentially the same but
operating at a lower frequency.
Frequency-hopping systems are the cheapest to make. Precise timing is needed to control the
frequency hops, but sophisticated signal processing is not required to extract the bit stream from the
radio signal. Direct-sequence systems require more sophisticated signal processing, which translates
into more specialized hardware and higher electrical power consumption. Direct-sequence techniques
also allow a higher data rate than frequency-hopping systems.
RF Propagation with 802.11
In fixed networks, signals are confined to wire pathways, so network engineers do not need to know
anything about the physics of electrical signal propagation. Instead, there are a few rules used to
calculate maximum cable length, and as long as the rules are obeyed, problems are rare. RF
propagation is not anywhere near as simple.
Signal Reception and Performance
Space is full of random electromagnetic waves, which can easily be heard by tuning a radio to an
unused frequency. Radio communication depends on making a signal intelligible over the background
noise. As conditions for reception degrade, the signal gets closer to the noise floor. Performance is
determined largely by the the most important factor, the signal-to-noise ratio (SNR). In Figure 10-2,
the SNR is illustrated by the height of the peak of the signal above the noise floor.
Figure 10-2. Signal-to-noise ratio and the noise floor
Having a strong signal is important, but not the whole story. Strong signals can be hard to pick out of
noisy environments. In some situations, it may be possible to raise the power to compensate for a
high noise floor. Unlicensed networks have only limited ability to raise power because of tight
regulatory constraints. As a result, more effort is placed on introducing as little additional noise as
possible before attempting to decode the radio signal.
The Shannon limit
Interestingly enough, there is no theoretical maximum to the amount of data that can be carried by a
radio channel. The capacity of a communications channel is given by the Shannon-Hartley theorem,
which was proved by Bell Labs researcher Claude Shannon in 1948. The theorem expresses the
mathematical limit of the capacity of a communications channel, which is named after the theorem
and is often called the Shannon limit or the Shannon capacity. The original Shannon theorem
expressed the maximum capacity C bits per second as a function of the bandwidth W in Hertz, and
the absolute ratio of signal power to noise. If the gain is measured in decibels, simply solve the
equation that defines a decibel for the power ratio for the second form.
Figure 10-3 shows the Shannon limit as a function of the signal-to-noise ratio. Shannon's theorem
reflects a theoretical reality of an unlimited bit rate. To get to an unlimited bit rate, the code designer
can require an arbitrarily large number of signal levels to distinguish between bits, but the fine
distinctions between the arbitrarily close signal rates will be swamped by the noise. One of the major
goals for 802.11 PHY designers is to design encoding rates as close to the Shannon limit as possible.
Alternatively, the Shannon theorem can be used to prescribe the minimum theoretical signal to noise
ratio to attain a given data rate. Solve the previous equations for the signal to noise ratio:
S/N = 2 ^ (C/W) - 1 (S/N as power ratio)
SNR = 10 * log10 (2 ^ (C/W) - 1) (SNR as dB)
Figure 10-3. Shannon limit as a function of SNR
Take 802.11a as an example. A single 20 MHz channel can carry a signal at a data rate up to 54
Mbps. Solving for the required signal-to-noise ratio yields 7.4 dB, which is much lower than what is
required by most real products on the market, reflecting the need of products to work in the realworld with much worse than ideal performance.[*]
[*]
Or, as another example, the telephone network makes channels with a frequency band of 180 Hz to 3.2 kHz, at a signal-tonoise ratio of 45 dB. That makes the maximum theoretical speed of an analog signal about 45 kbps. The trick behind the 56 kbps
modem is that the downstream path is all digital, and as a result has a higher signal-to-noise ratio.
Decibels and Signal Strength
Amplifiers may boost signals by orders of magnitude. Rather than keep track of all those
zeroes, amplifier power is measured in decibels (dB).
dB = 10 x log10 (power out/power in)
Decibel ratings are positive when the output is larger than the input and negative when
the output is smaller than the input. Each 10-dB change corresponds to a factor of 10,
and 3-dB changes are a factor of 2. Thus, a 33-dB change corresponds to a factor of
2000:
33 dB = 10 dB + 10 dB + 10 dB + 3 dB = 10 x 10 x 10 x 2 = 2000
Power is sometimes measured in dBm, which stands for dB above one milliwatt. To find
the dBm ratio, simply use 1 mW as the input power in the first equation.
It is helpful to remember that doubling the power is a 3-dB increase. A 1-dB increase is
roughly equivalent to a power increase of 25%. With these numbers in mind, you can
quickly perform most gain calculations in your head.
Path Loss, Range, and Throughput
In 802.11, the speed of the network depends on range. Several modulation types are defined by the
various 802.11 standards, ranging in speed from 1 Mbps to 54 Mbps. Receiver circuits must
distinguish between states to extract bits from radio waves. Higher speed modulations pack more
bits into a given time interval, and require a cleaner signal (and thus higher signal-to-noise ratio) to
successfully decode.
As radio signals travel through space, they degrade. For the most part, the noise floor will be
relatively constant over the limited range of an 802.11 network. Over distance, the degradation of
the signal will limit the signal-to-noise ratio at the receiver. As a station strays farther from an access
point, the signal level drops; with a constant noise floor, the degraded signal will result in a degraded
signal to noise ratio. Figure 10-4 illustrates this concept. As range from the access point increases,
the received signal gets closer to the noise floor. Stations that are closer have higher signal to noise
ratios. As a matter of network engineering, when the signal-to-noise ratio gets too small to support a
high data rate, the station will fall back to a lower data rate with less demanding signal-to-noise ratio
requirements.
Figure 10-4. Throughput versus distance
When there are no obstacles to obstruct the radio wave, the signal degradation can be calculated
with the following equation. The loss in free-space is sometimes called the path loss because it is the
minimum loss that would be expected along a path with a given length. Path loss depends on the
distance and the frequency of the radio wave. Higher distances and higher frequencies lead to higher
path loss. One reason why 802.11a has shorter range than 802.11b and 802.11g is that the path loss
is much higher at the 5 GHz used by 802.11a. The equation for free-space path loss is:
Path loss (dB) = 32.5 + 20 log F + log d
where the frequency F is expressed in GHz, and the distance d is expressed in meters. Path loss is
not, however, the only determinant of range. Obstacles such as walls and windows will reduce the
signal, and antennas and amplifiers may be used to boost the signal, which compensates for
transmission losses. Range calculations often include a fudge factor called the link margin to account
for unforeseen losses.
Total loss = TX power + TX antenna gain - path loss - obstacle loss - link margin + RX antenna
gain
Multipath Interference
Although there is a relatively simple equation for predicting radio propagation, it is only an estimate
for 802.11 networks. In addition to straight-line path losses, there are other phenomena that can
inhibit signal reception with 802.11. One of the major problems that plague radio networks is
multipath fading. Waves are added by superposition. When multiple waves converge on a point, the
total wave is simply the sum of any component waves. Figure 10-5 shows a few examples of
superposition.
Figure 10-5. Wave combination by superposition
In Figure 10-5 (c), the two waves are almost exactly the opposite of each other, so the net result is
almost nothing. Unfortunately, this result is more common than you might expect in wireless
networks. Most 802.11 equipment uses omnidirectional antennas, so RF energy is radiated in every
direction. Waves spread outward from the transmitting antenna in all directions and are reflected by
surfaces in the area. Figure 10-6 shows a highly simplified example of two stations in a rectangular
area with no obstructions.
Figure 10-6. Multiple paths
This figure shows three paths from the transmitter to the receiver. The wave at the receiver is the
sum of all the different components. It is certainly possible that the paths shown in Figure 10-6 will
all combine to give a net wave of 0, in which case the receiver will not understand the transmission
because there is no transmission to be received.
Because the interference is a delayed copy of the same transmission on a different path, the
phenomenon is called multipath fading or multipath interference. In many cases, multipath
interference can be resolved by changing the orientation or position of the receiver.
Inter-Symbol Interference (ISI)
Multipath fading is a special case of inter-symbol interference. Waves that take different paths from
the transmitter to the receiver will travel different distances and be delayed with respect to each
other, as in Figure 10-7. Once again, the two waves combine by superposition, but the effect is that
the total waveform is garbled. In real-world situations, wavefronts from multiple paths may be
added. The time between the arrival of the first wavefront and the last multipath echo is called the
delay spread. Longer delay spreads require more conservative coding mechanisms. 802.11b
networks can handle delay spreads of up to 500 ns, but performance is much better when the delay
spread is lower. When the delay spread is large, many cards will reduce the transmission rate;
several vendors claim that a 65-ns delay spread is required for full-speed 11-Mbps performance at a
reasonable frame error rate. (Discussion of reading the specification sheets is found in Chapter 16.)
Analysis tools can be used to measure the delay spread.
Figure 10-7. Inter-symbol interference
RF Engineering for 802.11
802.11 has been adopted at a stunning rate. Many network engineers accustomed to signals flowing
along well-defined cable paths are now faced with a LAN that runs over a noisy, error-prone, quirky
radio link. In data networking, the success of 802.11 has inexorably linked it with RF engineering. A
true introduction to RF engineering requires at least one book, and probably several. For the limited
purposes I have in mind, the massive topic of RF engineering can be divided into two parts: how to
make radio waves and how radio waves move.
RF Components
RF systems complement wired networks by extending them. Different components may be used
depending on the frequency and the distance that signals are required to reach, but all systems are
fundamentally the same and made from a relatively small number of distinct pieces. Two RF
components are of particular interest to 802.11 users: antennas and amplifiers. Antennas are of
general interest since they are the most tangible feature of an RF system. Amplifiers complement
antennas by allowing the antennas to pump out more power, which may be of interest depending on
the type of 802.11 network you are building.
Antennas
Antennas are the most critical component of any RF system because they convert electrical signals
on wires into radio waves and vice versa. In block diagrams, antennas are usually represented by a
triangular shape, as shown in Figure 10-8.
Figure 10-8. Antenna representations in diagrams
To function at all, an antenna must be made of conducting material. Radio waves hitting an antenna
cause electrons to flow in the conductor and create a current. Likewise, applying a current to an
antenna creates an electric field around the antenna. As the current to the antenna changes, so does
the electric field. A changing electric field causes a magnetic field, and the wave is off.
The size of the antenna you need depends on the frequency: the higher the frequency, the smaller
the antenna. The shortest simple antenna you can make at any frequency is 1/2 wavelength long
(although antenna engineers can play tricks to reduce antenna size further). This rule of thumb
accounts for the huge size of radio broadcast antennas and the small size of mobile phones. An AM
station broadcasting at 830 kHz has a wavelength of about 360 meters and a correspondingly large
antenna, but an 802.11b network interface operating in the 2.4-GHz band has a wavelength of just
12.5 centimeters. With some engineering tricks, an antenna can be incorporated into a PC Card or
around the laptop LCD screen, and a more effective external antenna can easily be carried in a
backpack or computer bag.
Antennas can also be designed with directional preference. Many antennas are omnidirectional, which
means they send and receive signals from any direction. Some applications may benefit from
directional antennas, which radiate and receive on a narrower portion of the field. Figure 10-9
compares the radiated power of omnidirectional and directional antennas.
Figure 10-9. Radiated power for omnidirectional and directional antennas
For a given amount of input power, a directional antenna can reach farther with a clearer signal.
They also have much higher sensitivity to radio signals in the dominant direction. When wireless links
are used to replace wireline networks, directional antennas are often used. Mobile telephone network
operators also use directional antennas when cells are subdivided. 802.11 networks typically use
omnidirectional antennas for both ends of the connection, although there are exceptionsparticularly if
you want the network to span a longer distance. Also, keep in mind that there is no such thing as a
truly omnidirectional antenna. We're accustomed to thinking of vertically mounted antennas as
omnidirectional because the signal doesn't vary significantly as you travel around the antenna in a
horizontal plane. But if you look at the signal radiated vertically (i.e., up or down) from the antenna,
you'll find that it's a different story. And that part of the story can become important if you're
building a network for a college or corporate campus and want to locate antennas on the top floors of
your buildings.
Of all the components presented in this section, antennas are the most likely to be separated from
the rest of the electronics. In this case, you need a transmission line (some kind of cable) between
the antenna and the transceiver. Transmission lines usually have an impedance of 50 ohms.
In terms of practical antennas for 802.11 devices in the 2.4-GHz band, the typical wireless PC Card
has an antenna built in. Built-in antennas work, but they will never be anything to write home about.
At best, the built in antenna in a PC card is mediocre. Larger antennas perform better. Some PC Card
802.11 interfaces have external antenna jacks. With an optional external antenna, the card has
better performance, at the cost of aesthetic quality. In response to the need for improved antenna
performance without ugly space-consuming external antennas, many laptops now use an antenna
built into the frame around the laptop screen.
Amplifiers
Amplifiers make signals bigger. Signal boost, or gain, is measured in decibels (dB). Amplifiers can be
broadly classified into three categories: low-noise, high-power, and everything else. Low-noise
amplifiers (LNAs) are usually connected to an antenna to boost the received signal to a level that is
recognizable by the electronics the RF system is connected to. LNAs are also rated for noise factor,
which is the measure of how much extraneous information the amplifier introduces. Smaller noise
factors allow the receiver to hear smaller signals and thus allow for a greater range.
High-power amplifiers (HPAs) are used to boost a signal to the maximum power possible before
transmission. Output power is measured in dBm, which are related to watts (see the "Decibels and
Signal Strength" sidebar earlier in this chapter). Amplifiers are subject to the laws of
thermodynamics, so they give off heat in addition to amplifying the signal. The transmitter in an
802.11 PC Card is necessarily low-power because it needs to run off a battery if it's installed in a
laptop, but it's possible to install an external amplifier at fixed access points, which can be connected
to the power grid where power is more plentiful.
This is where things can get tricky with respect to compliance with regulations. 802.11 devices are
limited to one watt of power output and four watts effective radiated power (ERP). ERP multiplies the
transmitter's power output by the gain of the antenna minus the loss in the transmission line. So if
you have a 1-watt amplifier, an antenna that gives you 8 dB of gain, and 2 dB of transmission line
loss, you have an ERP of 4 watts; the total system gain is 6 dB, which multiplies the transmitter's
power by a factor of 4.
Chapter 11. The Frequency-Hopping (FH)
PHY
Of all the physical layers standardized in the first draft of 802.11 in 1997, the frequency-hopping
spread spectrum (FH or FHSS) layer was the first layer to see widespread deployment. The
electronics used to support FH modulation are relatively cheap and do not have high power
requirements. Initially, the main advantage to using frequency-hopping networks was that a greater
number of networks could coexist, and the aggregate throughput of all the networks in a given area
was high. At this point, however, FH networks are largely a footnote in the history of 802.11.
Although it is a standard, only one vendor still manufactures and sells frequency-hopping systems,
and they are being phased out. Higher-throughput specifications have demolished the advantage of
aggregate throughput, and newer chipsets are less power-hungry.
This chapter describes the basic concepts used by the frequency-hopping PHY and the modulation
techniques used. It also shows how the physical layer convergence procedure prepares frames for
transmission on the radio link and touches briefly on a few details of the physical medium itself. At
this point, the FH PHY is largely a footnote in the history of 802.11, so you may want to skip this
chapter and move ahead to the next section on the direct-sequence PHY. However, understanding
how 802.11 technology developed will give you a better feeling for how all the pieces fit together.
Frequency-Hopping Transmission
Frequency hopping depends on rapidly changing the transmission frequency in a predetermined,
pseudorandom pattern, as illustrated in Figure 11-1. The vertical axis of the graph divides the
available frequency into a number of slots. Likewise, time is divided into a series of slots. A hopping
pattern controls how the slots are used. In the figure, the hopping pattern is {2,8,4,6}. Timing the
hops accurately is the key to success; both the transmitter and receiver must be synchronized so the
receiver is always listening on the transmitter's frequency.
Figure 11-1. Frequency hopping
Frequency hopping is similar to frequency division multiple access (FDMA) but with an important
twist. In FDMA systems, each device is allocated a fixed frequency. Multiple devices share the
available radio spectrum by using different frequencies. In frequency-hopping systems, the frequency
is time-dependent rather than fixed. Each frequency is used for a small amount of time, called the
dwell time.
Among other things, frequency hopping allows devices to avoid interfering with primary users
assigned to the same frequency band. It works because primary users are assigned narrow
frequency bands and the right to transmit at a power high enough to override the wireless LAN. Any
interference caused by the secondary user that affects the primary user is transient because the
hopping sequence spreads the energy out over a wide band.[*] Likewise, the primary user only
knocks out one of the spread-spectrum device's slots and looks like transient noise. Figure 11-2
shows the result when frequency slot 7 is given to a primary user. Although the transmission in the
fourth time slot is corrupted, the other three transmissions succeed.
[*]
If the primary user of a frequency band notices interference from secondary users, regulators can (and will) step in to shut down
the secondary user, hence the low power used by spread-spectrum modulation techniques.
If two frequency-hopping systems need to share the same band, they can be configured with
different hopping sequences so they do not interfere with each other. During each time slot, the two
hopping sequences must be on different frequency slots. As long as the systems stay on different
frequency slots, they do not interfere with each other, as shown in Figure 11-3. The gray rectangles
have a hopping sequence of {2,8,4,7}, as in the previous figures. A second system with a hopping
sequence of {6,3,7,2} is added. Hopping sequences that do not overlap are called orthogonal. When
multiple 802.11 networks are configured in a single area, orthogonal hopping sequences maximizes
throughput.
Figure 11-2. Avoiding interference with frequency hopping
Figure 11-3. Orthogonal hopping sequences
802.11 FH Details
802.11 divides the microwave ISM band into a series of 1-MHz channels. Approximately 99% of the
radio energy is confined to the channel. The modulation method used by 802.11 encodes data bits as
shifts in the transmission frequency from the channel center. Channels are defined by their center
frequencies, which begin at 2.400 GHz for channel 0. Successive channels are derived by adding 1MHz steps: channel 1 has a center frequency of 2.401 GHz, channel 2 has a center frequency of
2.402 GHz, and so on up to channel 95 at 2.495 GHz. Different regulatory authorities allow use of
different parts of the ISM band; the major regulatory domains and the available channels are shown
in Table 11-1.
Table 11-1. Channels used in different regulatory domains
Regulatory domain
Allowed channels
U.S. (FCC)
2 to 79 (2.402-2.479 GHz)
Canada (IC)
2 to 79 (2.402-2.479 GHz)
Europe (excluding France and Spain) (ETSI)
2 to 79 (2.402-2.479 GHz)
France
48 to 82 (2.448-2.482 GHz)
Spain
47 to 73 (2.447-2.473 GHz)
Japan (MKK)
73 to 95 (2.473-2.495 GHz)
The dwell time used by 802.11 FH systems is 390 time units, which is almost 0.4 seconds. When an
802.11 FH PHY hops between channels, the hopping process can take no longer than 224
microseconds. The frequency hops themselves are subject to extensive regulation, both in terms of
the size of each hop and the rate at which hops must occur.
802.11 Hop Sequences
Mathematical functions for deriving hop sets are part of the FH PHY specification and are found in
clause 14.6.8 of the 802.11 specification. As an example, hopping sequence 1 for North America and
most of Europe begins with the sequence {3, 26, 65, 11, 46, 19, 74, 50, 22, ...}. 802.11 further
divides hopping sequences into nonoverlapping sets, and any two members of a set are orthogonal
hopping sequences. In Europe and North America, each set contains 26 members. Regulatory
authorities in other areas have restricted the number of hopped channels, and therefore each set has
a smaller number of members. Table 11-2 has details.
Table 11-2. Size of hop sets in each regulatory domain
Regulatory domain
Hop set size
U.S. (FCC)
26
Canada (IC)
26
Europe (excluding France and Spain) (ETSI)
26
France
27
Spain
35
Japan (MIC)
23
Joining an 802.11 Frequency-Hopping Network
Joining a frequency-hopping network is made possible by the standardization of hop sequences.
Beacon frames on FH networks include a timestamp and the FH Parameter Set element. The FH
Parameter Set element includes the hop pattern number and a hop index. By receiving a Beacon
frame, a station knows everything it needs to synchronize its hopping pattern.
Based on the hop sequence number, the station knows the channel-hopping order. As an example,
say that a station has received a Beacon frame that indicates that the BSS is using the North
America/Europe hop sequence number 1 and is at hop index 2. By looking up the hop sequence, the
station can determine that the next channel is 65. Hop times are also well-defined. Each Beacon
frame includes a Timestamp field, and the hop occurs when the timestamp modulo dwell time
included in the Beacon is 0.
ISM Emission Rules and Maximum Throughput
Spectrum allocation policies are the limiting factor of frequency-hopping 802.11 systems. As an
example, consider the three major rules imposed by the FCC in the U.S.:[*]
[*]
These rules are in rule 247 of part 15 of the FCC rules (47 CFR 15.247).
1. There must be at least 75 hopping channels in the band, which is 83.5-MHz wide.
2. Hopping channels can be no wider than 1 MHz.
3. Devices must use all available channels equally. In a 30-second period, no more than 0.4
seconds may be spent using any one channel.
Of these rules, the most important is the second one. No matter what fancy encoding schemes are
available, only 1 MHz of bandwidth is available at any time. The frequency at which it is available
shifts continuously because of the other two rules, but the second rule limits the number of signal
transitions that can be used to encode data.
With a straightforward, two-level encoding, each cycle can encode one bit. At 1 bit per cycle, 1 MHz
yields a data rate of 1 Mbps. More sophisticated modulation and demodulation schemes can improve
the data rate. Four-level coding can pack 2 bits into a cycle, and 2 Mbps can be squeezed from the 1MHz bandwidth.
The European Telecommunications Standards Institute (ETSI) also has a set of rules for spreadspectrum devices in the ISM band, published in European Telecommunications Standard (ETS) 300328. The ETSI rules allow far fewer hopping channels; only 20 are required. Radiated power,
however, is controlled much more strictly. In practice, to meet both the FCC and ETSI requirements,
devices use the high number of hopping channels required by the FCC with the low radiated power
requirements of ETSI.
Effect of Interference
802.11 is a secondary use of the 2.4-GHz ISM band and must accept any interference from a higherpriority transmission. Catastrophic interference on a channel may prevent that channel from being
used but leave other channels unaffected. With approximately 80 usable channels in the U.S. and
Europe, interference on one channel reduces the raw bit rate of the medium by approximately
1.25%. (The cost at the IP layer will be somewhat higher because of the interframe gaps, 802.11
acknowledgments, and framing and physical-layer covergence headers.) As more channels are
affected by interference, the throughput continues to drop. See Figure 11-4.
Figure 11-4. Throughput response to interference in FHSS systems
Gaussian Frequency Shift Keying (GFSK)
The FH PHY uses Gaussian frequency shift keying (GFSK).[*] Frequency shift keying encodes data as
a series of frequency changes in a carrier. One advantage of using frequency to encode data is that
noise usually changes the amplitude of a signal; modulation systems that ignore amplitude
(broadcast FM radio, for example) tend to be relatively immune to noise. The Gaussian in GFSK
refers to the shape of radio pulses; GFSK confines emissions to a relatively narrow spectral band and
is thus appropriate for secondary uses. Signal processing techniques that prevent widespread
leakage of RF energy are a good thing, particularly for secondary users of a frequency band. By
reducing the potential for interference, GFSK makes it more likely that 802.11 wireless LANs can be
built in an area where another user has priority.
[*]
The term keying is a vestige of telegraphy. Transmission of data across telegraph lines required the use of a key. Sending data
through a modern digital system employs modulation techniques instead, but the word keying persists.
2-Level GFSK
The most basic GFSK implementation is called 2-level GFSK (2GFSK). Two different frequencies are
used, depending on whether the data that will be transmitted is a 1 or a 0. To transmit a 1, the
carrier frequency is increased by a certain deviation. Zero is encoded by decreasing the frequency by
the same deviation. Figure 11-5 illustrates the general procedure. In real-world systems, the
frequency deviations from the carrier are much smaller; the figure is deliberately exaggerated to
show how the encoding works.
The rate at which data is sent through the system is scalled the symbol rate. Because it takes several
cycles to determine the frequency of the underlying carrier and whether 1 or 0 was transmitted, the
symbol rate is a very small fraction of the carrier frequency. Although the carrier frequency is roughly
2.4 billion cycles per second, the symbol rate is only 1 or 2 million symbols per second.
Figure 11-5. 2-level GFSK
Frequency changes with GFSK are not sharp changes. Instantaneous frequency changes require more
expensive electronic components and higher power. Gradual frequency changes allow lower-cost
equipment with lower RF leakage. Figure 11-6 shows how frequency varies as a result of encoding
the letter M (1001101 binary) using 2GFSK. Note that the vertical axis is the frequency of the
transmission. When a 1 is transmitted, the frequency rises to the center frequency plus an offset,
and when a 0 is transmitted, the frequency drops by the same offset. The horizontal axis, which
represents time, is divided into symbol periods. Around the middle of each period, the receiver
measures the frequency of the transmission and translates that frequency into a symbol. (In 802.11
frequency-hopping systems, the higher-level data is scrambled before transmission, so the bit
sequence transmitted to the peer station is not the same as the bit sequence over the air. The figure
illustrates how the principles of 2GFSK work and doesn't step through an actual encoding.)
Figure 11-6. GFSK encoding of the letter M
4-Level GFSK
Using a scheme such as this, there are two ways to send more data: use a higher symbol rate or
encode more bits of information into each symbol. 4-level GFSK (4GFSK) uses the same basic
approach as 2GFSK but with four symbols instead of two. The four symbols (00, 01, 10, and 11) each
correspond to a discrete frequency, and therefore 4GFSK transmits twice as much data at the same
symbol rate. Obviously, this increase comes at a cost: 4GFSK requires more complex transmitters
and receivers. Mapping of the four symbols onto bits is shown in Figure 11-7.
Figure 11-7. Mapping of symbols to frequencies in 4GFSK
With its more sophisticated signal processing, 4GFSK packs multiple bits into a single symbol. Figure
11-8 shows how the letter M might be encoded. Once again, the vertical axis is frequency, and the
horizontal axis is divided into symbol times. The frequency changes to transmit the symbols; the
frequencies for each symbol are shown by the dashed lines. The figure also hints at the problem with
extending GFSK-based methods to higher bit rates. Distinguishing between two levels is fairly easy.
Four is harder. Each doubling of the bit rate requires that twice as many levels be present, and the
RF components distinguish between ever-smaller frequency changes. These limitations practically
limit the FH PHY to 2 Mbps.
Figure 11-8. 4GFSK encoding of the letter M
FH PHY Convergence Procedure (PLCP)
Before any frames can be modulated onto the RF carrier, the frames from the MAC must be prepared
by the Physical Layer Convergence Procedure (PLCP). Different underlying physical layers may have
different requirements, so 802.11 allows each physical layer some latitude in preparing MAC frames
for transmission over the air.
Framing and Whitening
The PLCP for the FH PHY adds a five-field header to the frame it receives from the MAC. The PLCP is a
relay between the MAC and the physical medium dependent (PMD) radio interface. In keeping with
ISO reference model terminology, frames passed from the MAC are PLCP service data units (PSDUs).
The PLCP framing is shown in Figure 11-9.
Preamble
As in a wired Ethernet, the preamble synchronizes the transmitter and receiver and derives
common timing relationships. In the 802.11 FH PHY, the Preamble is composed of the Sync
field and the Start Frame Delimiter field.
Figure 11-9. PLCP framing in the FH PHY
Sync
The sync field is 80 bits in length and is composed of an alternating zero-one sequence
(010101...01). Stations search for the sync pattern to prepare to receive data. In addition to
synchronizing the sender and receiver, the Sync field serves three purposes. First of all, the
presence of a sync signal indicates that a frame is imminent. Second, stations that have
multiple antennas to combat multipath fading or other environmental reception problems can
select the antenna with the strongest signal. Finally, the receiver can measure the frequency of
the incoming signal relative to its nominal values and perform any corrections needed to the
received signal.
Start Frame Delimiter (SFD)
As in Ethernet, the SFD signals the end of the preamble and marks the beginning of the frame.
The FH PHY uses a 16-bit SFD: 0000 1100 1011 1101.
Header
The PLCP header follows the preamble. The header has PHY-specific parameters used by the
PLCP. Three fields comprise the header: a length field, a speed field, and a frame check
sequence.
PSDU Length Word (PLW)
The first field in the PLCP header is the PLW. The payload of the PLCP frame is a MAC frame
that may be up to 4,095 bytes long. The 12-bit length field informs the receiver of the length
of the MAC frame that follows the PLCP header.
PLCP Signaling (PSF)
Bit 0, the first bit transmitted, is reserved and set to 0. Bits 1-3 encode the speed at which the
payload MAC frame is transmitted. Several speeds are available, so this field allows the
receiver to adjust to the appropriate demodulation scheme. Although the standard allows for
data rates in increments of 500 kbps from 1.0 Mbps to 4.5 Mbps, the modulation scheme has
been defined only for 1.0 Mbps and 2.0 Mbps.[*] See Table 11-3.
[*]
It is unlikely that significant further work will be done on high-rate, frequency-hopping systems. For
high data rates, direct sequence is a more cost-effective choice.
Table 11-3. PSF meaning
Bits (1-2-3)
Data rate
000
1.0 Mbps
001
1.5 Mbps
010
2.0 Mbps
011
2.5 Mbps
100
3.0 Mbps
101
3.5 Mbps
110
4.0 Mbps
111
4.5 Mbps
Header Error Check (HEC)
To protect against errors in the PLCP header, a 16-bit CRC is calculated over the contents of
the header and placed in this field. The header does not protect against errors in other parts of
the frame.
No restrictions are placed on the content of the Data field. Arbitrary data may contain long strings of
consecutive 0s or 1s, which makes the data much less random. To make the transmitted data more
like random white noise, the FH PHYs apply a whitening algorithm to the MAC frame. This algorithm
scrambles the data before radio transmission. Receivers invert the process to recover the data.
Frequency-Hopping PMD Sublayer
Although the PLCP header has a field for the speed at which the MAC frame is transmitted, only two
of these rates have corresponding standardized PMD layers. Several features are shared between
both PMDs: antenna diversity support, allowances for the ramp up and ramp down of the power
amplifiers in the antennas, and the use of a Gaussian pulse shaper to keep as much RF power as
possible in the narrow frequency-hopping band. Figure 11-10 shows the general design of the
transceiver used in 802.11 frequency-hopping networks. It is required that the transceiver have a
sensitivity of -80 dBm for both 1 Mbps and 2 Mbps transmission.
Figure 11-10. Frequency-hopping transceiver
PMD for 1.0-Mbps FH PHY
The basic frequency-hopping PMD enables data transmission at 1.0 Mbps. Frames from the MAC have
the PLCP header appended, and the resulting sequence of bits is transmitted out of the radio
interface. In keeping with the common regulatory restriction of a 1-MHz bandwidth, 1 million symbols
are transmitted per second. 2GFSK is used as the modulation scheme, so each symbol can be used
to encode a single bit. 802.11 specifies a minimum power of 10 milliwatts (mW) and requires the use
of a power control function to cap the radiated power at 100 mW, if necessary.
PMD for 2.0-Mbps FH PHY
A second, higher-speed PMD is available for the FH PHY. As with the 1.0-Mbps PMD, the PLCP header
is appended and is transmitted at 1.0 Mbps using 2GFSK. In the PLCP header, the PSF field indicates
the speed at which the frame body is transmitted. At the higher data rate, the frame body is
transmitted using a different encoding method than the physical-layer header. Regulatory
requirements restrict all PMDs to a symbol rate of 1 MHz, so 4GFSK must be used for the frame
body. Two bits per symbol yields a rate of 2.0 Mbps at 1 million symbols per second. Firmware that
supports the 2.0-Mbps PMD can fall back to the 1.0-Mbps PMD if signal quality is too poor to sustain
the higher rate.
Carrier sense/clear channel assessment (CS/CCA)
To implement the CSMA/CA foundation of 802.11, the PCLP includes a function to determine whether
the wireless medium is currently in use. The MAC uses both a virtual carrier-sense mechanism and a
physical carrier-sense mechanism; the physical layer implements the physical carrier sense. 802.11
does not specify how to determine whether a signal is present; vendors are free to innovate within
the required performance constraints of the standard. 802.11 requires that 802.11-compliant signals
with certain power levels must be detected with a corresponding minimum probability.
Characteristics of the FH PHY
Table 11-4 shows the values of a number of parameters in the FH PHY. In addition to the parameters
in the table, which are standardized, the FH PHY has a number of parameters that can be adjusted to
balance delays through various parts of an 802.11 frequency-hopping system. It includes variables
for the latency through the MAC, the PLCP, and the transceiver, as well as variables to account for
variations in the transceiver electronics. One other item of note is that the total aggregate
throughput of all frequency-hopping networks in an area can be quite high. The total aggregate
throughput is a function of the hop set size. All sequences in a hop set are orthogonal and
noninterfering. In North America and most of Europe, 26 frequency-hopping networks can be
deployed in an area at once. If each network is run at the optional 2 Mbps rate and half the airtime is
able to carry user payload data, the area can have a total of 26 Mbps throughput provided that the
ISM band is relatively free of interference.
Table 11-4. FH PHY parameters
Parameter
Value
Notes
Slot time
50ms
SIFS time
28ms
Contention
window size
15-1,023
slots
Preamble
duration
96ms
Preamble symbols are transmitted at 1 MHz, so a symbol takes 1 ms to
transmit; 96 bits require 96 symbol times.
PLCP header
duration
32ms
The PLCP header is 32 bits, so it requires 32 symbol times.
Maximum MAC
frame
4,095
bytes
802.11 recommends a maximum of 400 symbols (400 bytes at 1 Mbps,
800 bytes at 2 Mbps) to retain performance across different types of
environments.
Minimum
sensitivity
-80 dBm
The SIFS is used to derive the value of the other interframe spaces
(DIFS, PIFS, and EIFS).
Chapter 12. The Direct Sequence PHYs:
DSSS and HR/DSSS (802.11b)
The intial revision of the 802.11 specification in 1997 had a second physical layer based on direct
sequence spread spectrum (DSSS) technology. The DSSS PHY in 802.11 had data rates of 1 Mbps
and 2 Mbps. Although it operated at the same speed as the frequency hopping PHY, it quickly
became clear that direct sequence technolgies had the potential for higher speeds than frequency
hopping technologies. As a result, even though the two had equivalent speeds, direct sequence
became the PHY of choice. In 1999, a PHY with data rates of 5.5 Mbps and 11 Mbps was specified in
802.11b. The older 1 and 2 Mbps PHYs and the newer 5.5 and 11 Mbps PHYs are often combined into
a single interface, even though they are described by different specifications. (It is usually referred to
as "802.11b" support, even though the two lower rates are not part of 802.11b.) This chapter
describes the basic concepts and modulation techniques used by the direct sequence physical layers.
It also shows how the PLCP prepares frames for transmission on the radio link and touches briefly on
a few details of the physical medium itself.
Direct Sequence Transmission
Direct sequence transmission is an alternative spread-spectrum technique that can be used to
transmit a signal over a much wider frequency band. The basic approach of direct-sequence
techniques is to smear the RF energy over a wide band in a carefully controlled way. Changes in the
radio carrier are present across a wide band, and receivers can perform correlation processes to look
for changes. The basic high-level approach is shown in Figure 12-1.
Figure 12-1. Basic DSSS technique
At the left is a traditional narrowband radio signal. It is processed by a spreader, which applies a
mathematical transform to take a narrowband input and flatten the amplitude across a relatively
wide frequency band. To a narrowband receiver, the transmitted gnal looks like low-level noise
because its RF energy is spread across a very wide band. The key to direct-sequence transmission is
that any modulation of the RF carrier is also spread across the frequency band. Receivers can
monitor a wide frequency band and look for changes that occur across the entire band. The original
signal can be recovered with a correlator, which inverts the spreading process.
At a high level, a correlator simply looks for changes to the RF signal that occur across the entire
frequency band. Correlation gives direct-sequence transmissions a great deal of protection against
interference. Noise tends to take the form of relatively narrow pulses that, by definition, do not
produce coherent effects across the entire frequency band. Therefore, the correlation function
spreads out noise across the band, and the correlated signal shines through, as illustrated in Figure
12-2.
Figure 12-2. Spreading of noise by the correlation process
Direct-sequence modulation works by applying a chipping sequence to the data stream. A chip is a
binary digit used by the spreading process. Bits are higher-level data, while chips are binary numbers
used in the encoding process. There is no mathematical difference between a bit and a chip, but
spread-spectrum developers have adopted this terminology to indicate that chips are only a part of
the encoding and transmission process and do not carry any data. Chipping streams, which are also
called pseudorandom noise codes (PN codes), must run at a much higher rate than the underlying
data. (The requirement for high-speed oscillators to generate chip streams and recover data from
chip streams is one of the major power drains in direct sequence PHYs.) Figure 12-3 illustrates how
chipping sequences are used in the transmission of data using direct-sequence modulation. At the
left-hand side is a single data bit, which is either a one or a zero. For each data bit, several chips are
used. In the figure, the chip stream consists of an 11-bit code. It is combined with the single data
bit, to produce 11 chips that carry the single data bit. The 11-chip/single-data-bit sequence is
transmitted over the radio link and received at the other end. Every chunk of 11 chips is the
compared to an identical chip stream. If it matches the spreading code, a single zero data bit is
recovered; if not, a single one data bit is recovered.
The process of encoding a low bit-rate signal at a high chip rate has the side effect of spreading the
signal's power over a much wider bandwidth. One of the most important quantities in a directsequence system is its spreading ratio, which is the number
Figure 12-3. Chipping
of chips used to transmit a single bit.[*] Higher spreading ratios improve the ability to recover the
transmitted signal but require a higher chipping rate and a larger frequency band. Doubling the
spreading ratio requires doubling the chipping rate and doubles the required bandwidth as well.
There are two costs to increased chipping ratios. One is the direct cost of more expensive RF
components operating at the higher frequency, and the other is an indirect cost in the amount of
bandwidth required. Therefore, in designing direct-sequence systems for the real world, the
spreading ratio should be as low as possible to meet design requirements and to avoid wasting
bandwidth.
[*]
The spreading ratio is related to a figure known as the processing gain. The two are sometimes used interchangeably, but the
processing gain is slightly lower because it takes into account the effects of using real-world systems as opposed to perfect ideal
systems with no losses.
Direct-sequence modulation trades bandwidth for throughput. Compared to traditional narrowband
transmission, direct-sequence modulation requires significantly more radio spectrum and is much
slower. However, it can often coexist with other interference sources because the receiver's
correlation function effectively ignores narrowband noise. It is easier to achieve high throughput
using direct-sequence techniques than with frequency hopping. Regulatory authorities do not impose
a limit on the amount of spectrum that can be used; they generally set a minimum lower bound on
the processing gain. Higher rates can be achieved with a wider band, though wider bands require a
higher chip rate.
Encoding in 802.11 Direct Sequence Networks
For the PN code, 802.11 adopted an 11-bit Barker word. Each bit is encoded using the entire Barker
word as a chipping sequence. Detailed discussion of Barker words and their properties are well
beyond the scope of this book. The key attribute for 802.11 networks is that Barker words have good
autocorrelation properties, which means that the correlation function at the receiver operates as
expected in a wide range of environments and is relatively tolerant to multipath delay spreads. 11
bits are in the word because regulatory authorities often require a 10-dB processing gain in direct
sequence systems. Using an 11-bit spreading code for each bit allows 802.11 to meet the regulatory
requirements with some margin of safety, but it is small enough to allow as many overlapping
networks as possible. Longer spreading codes allow higher processing gains but require wider
frequency channels.
802.11 uses the Barker sequence {+1, -1, +1, +1, -1, +1, +1, +1, -1, -1, -1}. As used in 802.11,
+1 becomes 1, and -1 becomes 0, so the Barker sequence becomes 10110111000. It is applied to
each bit in the data stream by a modulo-2 adder.[*] When a 1 is encoded, all the bits in the
spreading code change; for 0, they stay the same. Figure 12-4 shows the encoding process.
[*]
Encoding with the Barker sequence is similar to a number of other techniques. Some cellular systems, most notably in North
America, use code division multiple access (CDMA) to allow several stations to access the radio medium. CDMA exploits some
extremely complex mathematics to ensure that transmissions from each mobile phone look like random noise to every other
mobile phone in the cell. The underlying mathematics are far more complicated than a simple, fixed pseudorandom noise code.
Figure 12-4. Encoding with the Barker word
Receivers can look at the number of 1s in a received bit time. The Barker sequence has six 1s and
five 0s. An 11-bit sequence with six 1s must therefore correspond to a transmitted 0, and an 11-bit
sequence with six 0s must correspond to a transmitted 1. In addition to counting the numbers of 1s
and 0s, the receiver can analyze the pattern of received bits to infer the value of the transmitted bit.
Radio Spectrum Usage in 802.11 Direct Sequence Networks
Channels for the DS PHY are much larger than the channels for the FH PHY. The DS PHY has 14
channels in the 2.4-GHz band, each 5 MHz wide. Channel 1 is placed at 2.412 GHz, channel 2 at
2.417 GHz, and so on up to channel 13 at 2.472 GHz. Channel 14 was defined especially for
operation in Japan, and has a center frequency that is 12 MHz from the center frequency of channel
13. Table 12-1 shows which channels are allowed by each regulatory authority. Channel 10 is
available throughout North America and Europe, which is why many products use channel 10 as the
default operating channel.
Table 12-1. Channels used in different regulatory domains
Regulatory domain
Allowed channels
U.S. (FCC)/Canada (IC)
1 to 11 (2.412-2.462 GHz)
Europe, excluding Spain (ETSI)
1 to 13 (2.412-2.472 GHz)
Spain
10 to 11 (2.457-2.462 GHz)
Japan (MIC)
1 to 13 (2.412-2.462 GHz) and 14 (2.484 GHz)
Channel energy spread
Within a channel, most of the energy is spread across a 22-MHz band. Because the DS PHY uses an
11-MHz chip clock, energy spreads out from the channel center in multiples of 11 MHz, as shown in
Figure 12-5. To prevent interference to adjacent channels, the first side lobe is filtered to 30 dB
below the power at the channel center frequency, and additional lobes are filtered to 50 dB below the
power at the channel center. This corresponds to reducing the power by a factor of 1,000 and
100,000, respectively. These limits are noted in Figure 12-5 by the use of dBr, which means dB
relative to the power at the channel center. Figure 12-5 uses a logarithmic scale: -30 dBr is only one
thousandth, and -50 dBr is one hundred thousandth.
Figure 12-5. Energy spread in a single 802.11 DS transmission channel
With the transmit filters in place, RF power is confined mostly to 22-MHz frequency bands. European
regulators cap the maximum radiated power at 100 mW; the FCC in the U.S. allows a substantially
higher transmission power of 1,000 mW. If directional antennas are used, the power can be
concentrated into an even higher radiated power.
Adjacent channel rejection and channel separation
To avoid interference, 802.11 equipment must be separated in the frequency domain. The initial
specification called for 35 dB rejection when two signals were 30 MHz apart, but the spectral
separation was tightened up to 25 MHz in 802.11b. To measure adjacent channel rejection, input a
maximum-speed signal at the center frequency of a channel, shown by the left curve in Figure 12-6.
Its power is 6 dB above the specified minimum sensitivity; for an 802.11b receiver, that is -70 dBm.
A second 802.11b signal at 11 Mbps is sent in, 35 dB stronger, but at least 25 MHz away. The right
curve depicts the interference test signal. As long as the frame error rate is less than 8%, the
receiver passes the test. Note that there is reasonably significant overlap between the first lobe of
the interfering signal and the main center lobe of the on-channel signal.
Figure 12-6. Adjacent channel rejection
Adjacent channel rejection also influences the number of channels that can be used simultaneously.
Although 802.11 networks operating in the ISM band have 14 defined channels, they are closely
packed. Real signals must be spaced out to prevent interference. 802.11b specifies that a 25 MHz (5
channel number) spacing is sufficient. Figure 12-7 shows the spectral mask of transmissions on the
so-called non-overlapping channels (1, 6, and 11). By comparing Figure 12-7 to the previous figure,
it is easy to see that using the nonoverlapping channel set requires all components in the system to
run at maximum effectiveness. Many real-world radios, however, are incapable of rejecting strong
adjacent-channel transmissions near the limit.
Although it is possible to have channels overlap more closely by using a four-channel layout (say, 1,
4, 7, and 11), it is not a generally useful deployment strategy. Greater overlap of the channels will
result in more interference between radios operating on adjacent frequencies. Greater interference
may trigger the carrier sensing mechanisms to report that the medium is busy or cause frames to be
mangled in flight. In either case, the peak throughput of each channel is less. It may be that trading
some amount of peak throughput for total area capacity is a good trade, but that is not generally
true.
Whether you call them 1, 6, and 11; A, B, and C; or Tom, Dick, and Harry, you
only have three mostly nonoverlapping channels with 2.4 GHz wireless LANs.
Figure 12-7. Channel separation in 802.11 DS networks
Maximum theoretical throughput
If the signal processing techniques used by the DS PHY are used, then the maximum throughput
would be a function of the frequency space used. Roughly speaking, the ISM band is 80-MHz wide.
Using the same spreading factor of 11 would lead to a maximum bit rate of slightly more than 7
Mbps. However, only one channel would be available, and products would need to have an oscillator
running at 77 MHz to generate the chipping sequence. High-frequency devices are a tremendous
drain on batteries, and the hypothetical high-rate encoding that uses the entire band makes terrible
use of the available spectrum. To achieve higher throughput, more sophisticated techniques must be
used. 802.11b increases the symbol rate slightly, but it gets far more mileage from more
sophisticated encoding techniques.
Interference response
Direct-sequence-modulated signals are more resistant to interference than frequency-hopping
signals. The correlation process enables direct-sequence systems to work around narrowband
interference much more effectively. With 11 chips per bit, several chips can be lost or damaged
before a single data bit is lost. The disadvantage is that the response of direct-sequence systems to
noise is not incremental. Up to a certain level, the correlator can remove noise, but once interference
obscures a certain amount of the frequency band, nothing can be recovered. Figure 12-8 shows how
direct-sequence systems degrade in response to noise.
Figure 12-8. Throughput response to interference in DSSS systems
Direct-sequence systems also avoid interfering with a primary user more effectively than frequencyhopping systems. After direct-sequence processing, signals are much wider and have lower
amplitudes, so they appear to be random background noise to traditional narrowband receivers. Two
direct-sequence users in the same area can cause problems for each other quite easily if the two
direct-sequence channels are not separated by an adequate amount. Generally speaking,
interference between two direct-sequence devices is a problem long before a primary band user
notices anything.
Differential Phase Shift Keying (DPSK)
Differential phase shift keying (DPSK) is the basis for 802.11 direct-sequence systems. As the name
implies, phase shift keying (PSK) encodes data in phase changes of the transmitted signal. The
absolute phase of a waveform is not relevant in PSK; only changes in the phase encode data. Like
frequency shift keying, PSK resists interference because most interference causes changes in
amplitude. Figure 12-8 shows two identical sine waves shifted by a small amount along the time axis.
The offset between the same point on two waves is the phase difference.
Figure 12-9. Phase difference between two sine waves
Differential Binary Phase Shift Keying (DBPSK)
The simplest form of PSK uses two carrier waves, shifted by a half cycle relative to each other. One
wave, the reference wave, is used to encode a 0; the half-cycle shifted wave is used to encode a 1.
Table 12-2 summarizes the phase shifts.
Table 12-2. DBPSK phase shifts
Symbol
Phase shift
0
0
1
180° (p radians)
Figure 12-10 illustrates the encoding as a phase difference from a preceding sine wave.
Figure 12-10. DBPSK encoding
To stick with the same example, encoding the letter M (1001101 in binary) is a matter of dividing up
the time into seven symbol times then transmitting the wave with appropriate phase shift at each
symbol boundary. Figure 12-11 illustrates the encoding. Time is divided into a series of symbol
periods, each of which is several times the period of the carrier wave. When the symbol is a 0, there
is no change from the phase of the previous symbol, and when the symbol is a 1, there is a change
of half a cycle. These changes result in "pinches" of the carrier when 1 is transmitted and a smooth
transition across the symbol time boundary for 0.
Figure 12-11. The letter M encoded in DBPSK
Differential Quadrature Phase Shift Keying (DQPSK)
Like 2GFSK, DBPSK is limited to one bit per symbol. More advanced receivers and transmitters can
encode multiple bits per symbol using a technique called differential quadrature phase shift keying
(DQPSK). Rather than a fundamental wave and a half-cycle shifted wave, DQPSK uses a fundamental
wave and three additional waves, each shifted by a quarter cycle, as shown in Figure 12-12. Table
12-3 summarizes the phase shifts.
Figure 12-12. DQPSK encoding
Table 12-3. DQPSK phase shifts
Symbol
Phase shift
00
0
01
90° (p/2 radians)
11
180° (p radians)
10
270° (3p/2 or -p/2 radians)
Now encode M in DQPSK (Figure 12-13). In the UTF-8 character set, M is represented by the binary
string 01001101 or, as the sequence of four two-bit symbols, 01-00-11-01. In the first symbol
period, there is a phase shift of 90 degrees; for clarity, the figure shows the phase shift from a pure
sine wave. The second symbol results in no phase shift, so the wave continues without a change. The
third symbol causes a phase shift of 180 degrees, as shown by the sharp change from the highest
amplitude to the lowest amplitude. The final symbol causes a phase shift of 90 degrees.
Figure 12-13. The letter M encoded in DQPSK
The obvious advantage of DQPSK relative to DBPSK is that the four-level encoding mechanism can
have a higher throughput. The cost of using DQPSK is that it cannot be used in some environments
because of severe multipath interference. Multipath interference occurs when the signal takes several
paths from the transmitter to the receiver. Each path has a different length; therefore, the received
signal from each path has a different delay relative to the other paths. This delay is the enemy of an
encoding scheme based on phase shifts. Wavefronts are not labeled or painted different colors, so a
wavefront could arrive later than expected because of a long path or it could simply have been
transmitted late and phase-shifted. In environments where multipath interference is severe, DQPSK
will break down much quicker than DBPSK.
The "Original" Direct Sequence PHY
The physical layer itself consists of two components. The Physical Layer Convergence Procedure
(PLCP) performs some additional PHY-dependent framing before transmission, while the Physical
Medium Dependent (PMD) layer is responsible for the actual transmission of frames.
PLCP Framing and Processing
The PLCP for the DS PHY adds a six-field header to the frames it receives from the MAC. In keeping
with ISO reference model terminology, frames passed from the MAC are PLCP service data units
(PSDUs). The PLCP framing is shown in Figure 12-14.
Figure 12-14. DS PLCP framing
The FH PHY uses a data whitener to randomize the data before transmission, but the data whitener
applies only to the MAC frame trailing the PLCP header. The DS PHY has a similar function called the
scrambler, but the scrambler is applied to the entirety of the direct-sequence frame, including the
PLCP header and preamble.
Preamble
The Preamble synchronizes the transmitter and receiver and allows them to derive common
timing relationships. It is composed of the Sync field and the Start Frame Delimiter field.
Before transmission, the preamble is scrambled using the direct-sequence scrambling function.
Sync
The Sync field is a 128-bit field composed entirely of 1s. Unlike the FH PHY, the Sync field is
scrambled before transmission.
Start Frame Delimiter (SFD)
The SFD allows the receiver to find the start of the frame, even if some of the sync bits were
lost in transit. This field is set to 0000 0101 1100 1111, which is different from the SFD used
by the FH PHY.
Header
The PLCP header follows the preamble. The header has PHY-specific parameters used by the
PLCP. Five fields comprise the header: a signaling field, a service identification field, a Length
field, a Signal field used to encode the speed, and a frame-check sequence.
Signal
The Signal field is used by the receiver to identify the transmission rate of the encapsulated
MAC frame. It is set to either 0000 1010 (0x0A) for 1-Mbps operation or 0001 0100 (0x14) for
2-Mbps operation.
Service
This field is reserved for future use and must be set to all 0s.
Length
This field is set to the number of microseconds required to transmit the frame as an unsigned
16-bit integer, transmitted least-significant bit to most-significant bit.
CRC
To protect the header against corruption on the radio link, the sender calculates a 16-bit CRC
over the contents of the four header fields. Receivers verify the CRC before further frame
processing.
No restrictions are placed on the content of the Data field. Arbitrary data may contain long strings of
consecutive 0s or 1s, which makes the data much less random. To make the data more like random
background noise, the DS PHY uses a polynomial scrambling mechanism to remove long strings of 1s
or 0s from the transmitted data stream.
DS Physical Medium Dependent Sublayer
The PMD is a complex and lengthy specification that incorporates provisions for two data rates (1.0
and 2.0 Mbps). Figure 12-15 shows the general design of a transceiver for 802.11 direct-sequence
networks.
Figure 12-15. Direct-sequence transceiver
Transmission at 1.0 Mbps
At the low data rate, the direct-sequence PMD enables data transmission at 1.0 Mbps. The PLCP
header is appended to frames arriving from the MAC, and the entire unit is scrambled. The resulting
sequence of bits is transmitted from the physical interface using DBPSK at a rate of 1 million symbols
per second. The resulting throughput is 1.0 Mbps because one bit is encoded per symbol. Like the FH
PMD, the DS PMD has a minimum power requirement and can cap the power at 100 mW if necessary
to meet regulatory requirements.
Transmission at 2.0 Mbps
Like the FH PHY, transmission at 2.0 Mbps uses two encoding schemes. The PLCP preamble and
header are transmitted at 1.0 Mbps using DBPSK. Although using a slower method for the header
transmission reduces the effective throughput, DBPSK is far more tolerant of noise and multipath
interference. After the preamble and header are finished, the PMD switches to DQPSK modulation to
provide 2.0-Mbps service. As with the FH PHY, most products that implement the 2.0-Mbps rate can
detect interference and fall back to lower-speed 1.0-Mbps service.
CS/CCA for the DS PHY
802.11 allows the carrier sense/clear channel assessment function to operate in one of three modes:
Mode 1
When the energy exceeds the energy detection (ED) threshold, it reports that the medium is
busy. The ED threshold depends on the transmit power.
Mode 2
Implementations using Mode 2 must look for an actual DSSS signal and report the channel
busy when one is detected, even if the signal is below the ED threshold.
Mode 3
Mode 3 combines Mode 1 and Mode 2. A signal must be detected with sufficient energy before
the channel is reported busy to higher layers.
Once a channel is reported busy, it stays busy for the duration of the intended transmission, even if
the signal is lost. The transmission's duration is taken from the time interval in the Length field. Busy
medium reports must be very fast. When a signal is detected at the beginning of a contention
window slot, the CCA mechanism must report a busy medium by the time the slot has ended. This
relatively high performance requirement must be set because once a station has begun transmission
at the end of its contention delay, it should seize the medium, and all other stations should defer
access until its frame has concluded.
Characteristics of the DS PHY
Table 12-4 shows the values of a number of parameters in the DS PHY. In addition to the parameters
in the table, which are standardized, the DS PHY has a number of parameters that can be adjusted to
balance delays through various parts of an 802.11 direct-sequence system. It includes variables for
the latency through the MAC, the PLCP, and the transceiver, as well as variables to account for
variations in the transceiver electronics. One other item of note is that the total aggregate
throughput of all direct-sequence networks in an area is much lower than the total aggregate
throughput of all nonoverlapping frequency-hopping networks in an area. The total aggregate
throughput is a function of the number of nonoverlapping channels. In North America and most of
Europe, three direct-sequence networks can be deployed in an area at once. If each network is run at
the optional 2 Mbps rate and the efficiency of the protocol allows 50% of the headline rate to become
user data throughput, the total throughput is 3 Mbps, which is dramatically less than the frequencyhopping total aggregate throughput.
Table 12-4. DS PHY parameters
Parameter
Value
Notes
Slot time
20 ms
SIFS time
10 ms
Contention window
size
31 to 1,023
slots
Preamble duration
144 ms
Preamble symbols are transmitted at 1 MHz, so a symbol takes 1
ms to transmit; 144 bits require 144 symbol times.
PLCP header
duration
48 ms
The PLCP header is 48 bits, so it requires 48 symbol times.
Maximum MAC
frame
4-8,191
bytes
Minimum receiver
sensitivity
-80 dBm
Adjacent channel
rejection
35 dB
The SIFS is used to derive the value of the other interframe
spaces (DIFS, PIFS, and EIFS).
See text for measurement details.
Like the FH PHY, the DS PHY has a number of attributes that can be adjusted by a vendor to balance
delays in various parts of the system. It includes variables for the latency through the MAC, the
PLCP, and the transceiver, as well as variables to account for variations in the transceiver electronics.
Complementary Code Keying
802.11 direct-sequence systems use a rate of 11 million chips per second. The original DS PHYs
divided the chip stream up into a series of 11-bit Barker words and transmitted 1 million Barker
words per second. Each word encoded either one bit or two bits for a corresponding data rate of 1.0
Mbps or 2.0 Mbps, respectively. Achieving higher data rates and commercial utility requires that each
code symbol carry more information than a bit or two.
Straight phase shift encoding cannot hope to carry more than a few bits per code word. DQPSK
requires that receivers distinguish quarter-cycle phase differences. Further increasing the number of
bits per symbol would require processing even finer phase shifts, such as an eighth-cycle or
sixteenth-cycle shift. Detecting smaller phase shifts is more difficult in the presence of multipath
interference and requires more sophisticated (and thus expensive) electronics.
Instead of continuing with straight phase-shift keying, the IEEE 802.11 working group turned to an
alternate encoding method. Complementary code keying (CCK) divides the chip stream into a series
of 8-bit code symbols, so the underlying transmission is based on a series of 1.375 million code
symbols per second. CCK is based on sophisticated mathematical transforms that allow the use of a
few 8-bit sequences to encode 4 or even 8 bits per code word, for a data throughput of 5.5 Mbps or
11 Mbps. In addition, the mathematics underlying CCK transforms allow receivers to distinguish
between different codes easily, even in the presence of interference and multipath fading. Figure 1216 illustrates the use of code symbols in CCK. It is quite similar to the chipping process used by the
slower direct-sequence layers; the difference is that the code words are derived partially from the
data. A static repeating code word such as the Barker word is not used.
Figure 12-16. Code symbols in CCK
Barker spreading, as used in the lower-rate, direct-sequence layers, uses a static code to spread the
signal over the available frequency band. CCK uses the code word to carry information, as well as
simply to spread the signal. Several phase angles are used to prepare a complex code word of eight
bits.
High Rate Direct Sequence PHY
To distinguish it from the original direct sequence PHY, the high rate PHY that runs at 11 Mbps is
abbreviated as HR/DSSS. Like its predecessor, it is split into a convergence procedure that prepares
frames for radio transmission, and a medium-dependent layer that turns the bits into radio waves in
the air.
PLCP Framing and Scrambling
The long headers required by the original PHY greatly reduce performance. The 802.11 MAC requires
an acknowledgment for every data frame, and the 192 microsecond preamble is much, much longer
than the MAC acknowledgment. At the 11 Mbps data rate, the preamble and PLCP framing header
sucks up 25% of the time used to transmit a 1,500 byte frame and its corresponding MAC
acknowledgment. As long as a new PHY was being developed, the designers of 802.11b came up with
a new "short" framing format that improves protocol efficiency and, hence, throughput. Using the
short headers cuts the preamble and PLCP framing overhead cuts the preamble and framing
overhead to 14%. While still substantial, it is a dramatic improvement. Figure 12-17 shows the PLCP
framing specified in 802.11b. When 802.11b was first released, short headers were not supported by
all devices because of the large installed base of 2 Mbps direct sequence equipment. At this point,
almost every card can safely support the short preamble, and most access point vendors use it by
default.
Naturally, the optional short format may be used only if all stations support it. To prevent networks
configured for the short format from disappearing, 802.11b requires that stations answering Probe
Requests from an active scan return a response using the same PLCP header that was received. If a
station that supports
Figure 12-17. HR/DSSS PLCP framing
only the long PLCP header sends a Probe Response, an access point returns a response using the
long header, even if the BSS is configured for the short header.
Preamble
Frames begin with the preamble, which is composed of the Sync field and the SFD field. The
preamble is transmitted at 1.0 Mbps using DBPSK.
Long Sync
The Long Sync field is composed of 128 one bits. It is processed by the scrambler before
transmission, though, so the data content varies. High-rate systems use a specified seed for
the scrambling function but support backwards compatibility with older systems that do not
specify a seed.
Short Sync
The Short Sync field is composed of 56 zero bits. Like the Long Sync, it is also processed by
the scrambler.
Long SFD
To indicate the end of the Sync field, the long preamble concludes with a Start of Frame
Delimiter (SFD). In the long PLCP, the SFD is the sequence 1111 0011 1010 0000. As with all
IEEE specifications, the order of transmission from the physical interface is least-significant bit
first, so the string is transmitted right to left.
Short SFD
To avoid confusion with the Long SFD, the Short SFD is the reverse value, 0000 0101 1100
1111.
The PLCP header follows the preamble. It is composed of the Signal, Service, Length, and CRC fields.
The long header is transmitted at 1.0 Mbps using DBPSK. However, the short header's purpose is to
reduce the time required for overhead transmission so it is transmitted at 2.0 Mbps using DQPSK.
Long Signal
The Long Signal field indicates the speed and transmission method of the enclosed MAC frame.
Four values for the 8-bit code are currently defined and are shown in Table 12-5.
Table 12-5. Signal field values
Speed
Value (msb to lsb)
Hex value
1 Mbps
0000 1010
0x0A
2 Mbps
0001 0100
0x14
5.5 Mbps
0011 0111
0x37
11 Mbps
0110 1110
0x6E
Short Signal
The Short Signal field indicates the speed and transmission method of the enclosed frame, but
only three values are defined. Short preambles can be used only with 2 Mbps, 5.5 Mbps, and
11 Mbps networks.
Service
The Service field, which is shown in Figure 12-18, was reserved for future use by the first
version of 802.11, and bits were promptly used for the high-rate extensions in 802.11b. First
of all, the Length field describes the amount of time used for the enclosed frame in
microseconds. Above 8 Mbps, the value becomes ambiguous. Therefore, the eighth bit of the
service field is used to extend the Length field to 17 bits. The third bit indicates whether the
802.11b implementation uses locked clocks; clock locking means that transmit frequency and
symbol clock use the same oscillator. The fourth bit indicates the type of coding used for the
packet, which is either 0 for CCK or 1 for PBCC. All reserved bits must be set to 0. The Service
field is transmitted from left to right (b0 to b7), which is the same in both the short and long
PLCP frame formats. (Further changes are made by 802.11g, which will be discussed in
Chapter 14.)
Figure 12-18. Service field in the HR/DSSS PLCP header
Length
The Length field is the same in both the short and long PLCP frame formats and is the number
of microseconds required to transmit the enclosed MAC frame. Approximately two pages of the
802.11b standard are devoted to calculating the value of the Length frame, but the details are
beyond the scope of this book.
CRC
The CRC field is the same in both the short and the long PLCP frames. Senders calculate a CRC
checksum using the Signal, Service, and Length fields. Receivers can use the CRC value to
ensure that the header was received intact and was not damaged during transmission. CRC
calculations take place before data scrambling.
The data scrambling procedure for the HR/DSSS PHY is nearly identical to the data scrambling
procedure used with the original DS PHY. The only difference is that the scrambling function is
seeded to specified values in the HR/DSSS PHY. Different seeds are used for short and long PLCP
frames.
HR/DSSS PMD
Like the DS PHY, the 802.11b PHY uses a single PMD specification. The general transceiver design is
shown in Figure 12-19.
Figure 12-19. HR/DSSS transceiver
Transmission at 1.0 Mbps or 2.0 Mbps
To ensure backwards compatibility with the installed base of 802.11-based, direct-sequence
hardware, the HR/DSSS PHY can transmit and receive at 1.0 Mbps or 2.0 Mbps. Slower transmissions
are supported in the same manner as the lower-rate, direct-sequence layers described in previously
in this chapter. Any transmissions at the slower rates must use long headers.
Transmission at 5.5 Mbps with CCK
Higher-rate transmission is accomplished by building on the DQPSK-based phase shift keying
techniques. DQPSK transmits two bits per symbol period, encoded as one of four different phase
shifts. By using CCK, the symbol words themselves carry additional information. 5.5-Mbps
transmission encodes four data bits into a symbol. Two bits are carried using conventional DQPSK,
and the other two are carried through the content of the code words. Figure 12-20 illustrates the
overall process.
Figure 12-20. 802.11b transmission at 5.5 Mbps
1.
1. The MAC frame embedded in the PCLP frame is divided into a string of 4-bit blocks. Each 4-bit
block is further divided into two 2-bit segments.
2. The first 2-bit segment is encoded by means of a DQPSK-type phase shift between the current
symbol and the previous symbol (Table 12-6). Even and odd symbols use a different phase shift
for technical reasons. Symbol numbering starts with 0 for the first 4- bit block.
Table 12-6. Inter-symbol DQPSK phase shifts
Bit pattern
Phase angle (even symbols)
Phase angle (odd symbols)
00
0
P
01
p/2
3p/2
11
p
0
10
3p/2
p/2
3. The second 2-bit segment is used to select one of four code words for the current symbol (Table
12-7). The four code words can be derived using the mathematics laid out in clause 18.4.6.5 of
the 802.11 standard.
Table 12-7. Mbps code words
Bit sequence
Code word
00
i,1,i,-1,i,1,-i,1
01
-i,-1,-i,1,1,1,-i,1
10
-i,1,-i,-1,-i,1,i,1
11
i,-1,i,1,-i,1,i,1
Transmission at 11 Mbps with CCK
To move to a full 11 Mbps, 8 bits must be encoded with each symbol. As with other techniques, the
first two bits are encoded by the phase shift of the transmitted symbol relative to the previous
symbol. Six bits are encoded using CCK. Figure 12-21 illustrates the process.
Figure 12-21. 802.11b transmission at 11 Mbps
1. The MAC frame embedded in the PCLP frame is divided into a string of 8-bit blocks. Each 8-bit
block is further divided into four 2-bit segments.
2. The first 2-bit segment is encoded by means of a DQPSK-type phase shift between the current
symbol and the previous symbol. As with the 5.5-Mbps rate, even and odd symbols use a
different phase shift for technical reasons. Symbol numbering starts with 0 for the first 8-bit
block. The phase shifts are identical to the phase shifts used in 5.5-Mbps transmission.
3. The remaining six bits are grouped into three successive pairs. Each pair is associated with the
phase angle in Table 12-8 and is used to derive a code word.
Table 12-8. Phase angle encoding for 11-Mbps transmission
Bit pattern
Phase angle
00
0
01
p/2
10
p
11
3p/2
As an example, consider the conversion of the bit sequence 0100 1101 into a complex code for
transmission on an 802.11b network. The first two bits, 01, encode a phase shift from the previous
symbol. If the symbol is an even symbol in the MAC frame, the phase shift is p/2; otherwise, the shift
is 3p/2. (Symbols in the MAC frame are numbered starting from 0, so the first symbol in a frame is
even.) The last six bits are divided into three 2-bit groups: 00, 11, and 01. Each of these is used to
encode an angle in the code word equation. The next step in transmission is to convert the phase
angles into the complex code word for transmission.
Clear channel assessment
Like the original DS PHY, high-rate implementers have three choices for the CS/CCA operation mode.
All the direct-sequence CCA modes are considered to be part of the same list. Mode 1 is identical to
the DS PHY's CCA Mode 1, and Modes 2 and 3 are used exclusively by the original DS PHY. Modes 4
and 5 are the HR/DSSS-specific CCA modes.
Mode 1
When the energy exceeds the energy detection (ED) threshold, the medium is reported busy.
The ED threshold depends on the transmit power used. This mode is also available for classic
direct-sequence systems.
Mode 4
Implementations using Mode 4 look for an actual signal. When triggered, a Mode 4 CCA
implementation starts a 3.65 ms timer and begins counting down. If no valid HR/DSSS signal is
received by the expiration of the timer, the medium is reported idle. 3.65 ms corresponds to
the transmission time required for the largest possible frame at 5.5 Mbps.
Mode 5
Mode 5 combines Mode 1 and Mode 4. A signal must be detected with sufficient energy before
the channel is reported busy to higher layers.
Once a channel is reported busy, it stays busy for the duration of the intended transmission, even if
the signal is lost. The channel is considered busy until the time interval in the Length field has
elapsed. Implementations that look for a valid signal may override this requirement if a second PLCP
header is detected.
Optional Features of the 802.11b PHY
802.11b includes two optional physical-layer features, neither of which is widely used. Packet Binary
Convolutional Coding (PBCC) was proposed as a method of reaching the 11 Mbps data rate, but it
was never widely implemented. Proposals for further revisions to wireless LAN technology in the ISM
band specified PBCC, but those proposals were rejected in the summer of 2001.
A second optional feature, channel agility, was designed to assist networks in avoiding interference.
Channel agility causes the center channel to shift periodically in the hope that interference can be
avoided. Channel agility was never widely used because it was not particularly helpful. In the
presence of interference, some throughput would be recovered as receivers hopped to another
channel, but the additional spectrum required made it much more effective to hunt down and fix the
interference, or remap channels around it.
Characteristics of the HR/DSSS PHY
Table 12-9 shows the values of a number of parameters in the HR/DSSS PHY. Like the DS PHY, the
HR/DSSS PHY has a number of parameters that can be adjusted to compensate for delays in any part
of a real system.
Table 12-9. HR/DSSS PHY parameters
Parameter
Value
Notes
Maximum MAC
frame length
4,095 bytes
Slot time
20 ms
SIFS time
10 ms
Contention window
size
31 to 1,023
slots
Preamble duration
144 ms
Preamble symbols are transmitted at 1 MHz, so a symbol takes 1
ms to transmit; 96 bits require 96 symbol times.
PLCP header
duration
48 bits
The PLCP header transmission time depends on whether the short
preamble is used.
The SIFS is used to derive the value of the other interframe
spaces (DIFS, PIFS, and EIFS).
Minimum sensitivity -76 dBm
Adjacent channel
rejection
35 dB
See text for measurement notes.
One other item of note is that the total aggregate throughput of all HR/DSSS networks in an area is
still lower than the total aggregate throughput of all nonoverlapping frequency-hopping networks in
an area. The total aggregate throughput is a function of the number of nonoverlapping channels. In
North America and most of Europe, three HR/DSSS networks can be deployed in an area at once.
Running each network at the top speed of 11 Mbps, and assuming user payload data throughput of
50%, the total aggregate throughput will be 16.5 Mbps.
Chapter 13. 802.11a and 802.11j: 5-GHz
OFDM PHY
Go West, young man, and grow up with the country!
John B.L. Soule[*]
[*]
This is often attributed to Horace Greeley, who did much to popularize it.
The 2.4 GHz ISM bands are crowded, and often choked with non-802.11 traffic. In an attempt to
develop higher data rates, the 802.11 working group standardized a physical layer using unlicensed
spectrum around 5 GHz. Large tracts of undeveloped spectrum were available for unlicensed use,
and far fewer devices have used the spectrum.
802.11a was standardized in 1999, but it took a significant amount of time to bring products to
market. 802.11a hardware finally hit the market in late 2001. Today, the best known manufacturer
of 802.11a chipsets is Atheros Communications. In form and function, they look similar to any other
wireless LAN card. Most are CardBus cards, although some additional form factors are used more
widely now. The range of very high data rates is quite short, but generally, 802.11a has a
comparable data rate to 802.11b at comparable distance.
Initially, 802.11a was designed for the 5-GHz Unlicensed National Information Infrastructure (U-NII)
bands in the United States. With the success of 802.11a in the American market, other regulators
developed rules to allow 802.11a. The 802.11 working group supported these efforts through
802.11h for Europe, which was discussed in Chapter 8, and 802.11j for Japan.
This chapter begins with a qualitative introduction to the basis of OFDM. When all the mathematical
formalism is stripped away, OFDM is a method for chopping a large frequency channel into a number
of subchannels. The subchannels are then used in parallel for higher throughput. I anticipate that
many readers will skip the first section, either because they are already familiar with OFDM or are
interested only in how the frequency bands are used and how the PCLP wraps up frames for
transmission.
Orthogonal Frequency Division Multiplexing (OFDM)
802.11a is based on orthogonal frequency division multiplexing (OFDM). OFDM is not a new
technique. Most of the fundamental work was done in the late 1960s, and U.S. patent number
3,488,445 was issued in January 1970. Recent DSL work (HDSL, VDSL, and ADSL) and wireless data
applications have rekindled interest in OFDM, especially now that better signal-processing techniques
make it more practical.[*] OFDM does, however, differ from other emerging encoding techniques such
as code division multiple access (CDMA) in its approach. CDMA uses complex mathematical
transforms to put multiple transmissions onto a single carrier; OFDM encodes a single transmission
into multiple subcarriers. The mathematics underlying the code division in CDMA is far more
complicated than in OFDM.
[*]
The lack of interest in OFDM means that references on it are sparse. Readers interested in the mathematical background that
is omitted in this chapter should consult OFDM for Wireless Multimedia Applications by Richard van Nee and Ramjee Prasad
(Artech House, 2000).
OFDM devices use one wide-frequency channel by breaking it up into several component
subchannels. Each subchannel is used to transmit data. All the "slow" subchannels are then
multiplexed into one "fast" combined channel.
Carrier Multiplexing
When network managers solicit user input on network build-outs, one of the most common demands
is for more speed. The hunger for increased data transmissions has driven a host of technologies to
increase speed. OFDM takes a qualitatively similar approach to Multilink PPP: when one link isn't
enough, use several in parallel.
OFDM is closely related to plain old frequency division multiplexing (FDM). Both divide the available
bandwidth into slices called carriers or subcarriers and make those carriers available as distinct
channels for data transmission. OFDM boosts throughput by using several subcarriers in parallel and
multiplexing data over the set of subcarriers.
Traditional FDM was widely used by first-generation mobile telephones as a method for radio channel
allocation. Each user was given an exclusive channel, and guard bands were used to ensure that
spectral leakage from one user did not cause problems for users of adjacent channels. Figure 13-1
illustrates the traditional FDM approach.
The trouble with traditional FDM is that the guard bands waste bandwidth and thus reduce capacity.
To avoid wasting transmission capacity with unused guard bands, OFDM selects channels that
overlap but do not interfere with each other. Figure 13-2 illustrates the contrast between traditional
FDM and OFDM.
Figure 13-1. Traditional FDM
Figure 13-2. FDM versus OFDM
Overlapping carriers are allowed because the subcarriers are defined so that they are easily
distinguished from one another. The ability to separate the subcarriers hinges on a complex
mathematical relationship called orthogonality.
Orthogonality Explained (Without Calculus)
Orthogonal is a mathematical term derived from the Greek word orthos, meaning straight, right, or
true. In mathematics, the word "orthogonal" is used to describe independent items. Orthogonality is
best seen in the frequency domain, looking at a spectral breakdown of a signal. OFDM works because
the frequencies of the subcarriers are selected so that at each subcarrier frequency, all other
subcarriers do not contribute to the overall waveform. One common way of looking at orthogonality
is shown in Figure 13-3. The signal has been divided into its three subcarriers. The peak of each
subcarrier, shown by the heavy dot at the top, encodes data. The subcarrier set is carefully designed
to be orthogonal; note that at the peak of each of the subcarriers, the other two subcarriers have
zero amplitude.
OFDM takes the coded signal for each subchannel and uses the inverse fast Fourier transform (IFFT)
to create a composite waveform from the strength of each subchannel. OFDM receivers can then
apply the FFT to a received waveform to extract the amplitude of each component subcarrier.
Guard Time
With the physical layers discussed in Chapter 12, the main problem for receivers was inter-symbol
interference (ISI) (Figure 13-4). ISI occurs when the delay spread between different paths is large
and causes a delayed copy of the transmitted bits to shift onto a previously arrived copy.
Figure 13-3. Orthogonality in the frequency domain
Figure 13-4. ISI reviewed
With OFDM, inter-symbol interference does not pose the same kind of problem. The Fourier
transform used by OFDM distills the received waveform into the strengths of the subcarriers, so time
shifts do not cause dramatic problems. In Figure 13-4, there would be a strong peak for the
fundamental low-frequency carrier, and the late-arriving high-frequency component could be
ignored.
As with all benefits, however, there is a price to pay. OFDM systems use multiple subcarriers of
different frequencies. The subcarriers are packed tightly into an operating channel, and small shifts in
subcarrier frequencies may cause interference between carriers, a phenomenon called inter-carrier
interference (ICI). Frequency shifts may occur because of the Doppler effect or because there is a
slight difference between the transmitter and receiver clock frequencies.
Fourier Analysis, the Fourier Transform, and Signal Processing
The Fourier transform is often called "the Swiss Army knife of signal processing." Signal
processing often defines actions in terms of frequency components. Receivers, however,
process a time-varying signal amplitude. The Fourier transform is a mathematical
operation that divides a waveform into its component parts. Fourier analysis takes a
time-varying signal and converts it to the set of frequency-domain components that
make up the signal.
Signal-processing applications often need to perform the reverse operation as well. Given
a set of frequency components, these applications use them like a recipe to build the
composite waveform. The mathematical operation used to build the composite waveform
from the known ingredients in the frequency domain is the inverse Fourier transform.
Strictly speaking, Fourier analysis is applied to smooth curves of the sort found in
physics textbooks. To work with a set of discrete data points, a relative of Fourier
transform called the discrete Fourier transform (DFT) must be used. Like the Fourier
transform, the DFT has its inverted partner, the inverse DFT (IDFT).
The DFT is a computationally intensive process of order N2, which means that its running
time is proportional to the square of the size of the number of data points. If the number
of data points is an even power of two, however, several computational shortcuts can be
taken to cut the complexity to order N log N. On large data sets, the reduced complexity
boosts the speed of the algorithm. As a result, the "short" DFT applied to 2 n data points
is called the fast Fourier transform (FFT). It also has an inverted relative, the inverse fast
Fourier transform (IFFT).
Fast Fourier transforms used to be the domain of supercomputers or special-purpose,
signal-processing hardware. But with the microprocessor speeds available today,
sophisticated signal processing is well within the capabilities of a PC. Specialized digital
signal processors (DSPs) are now cheap enough to be used in almost anythingincluding
the chip sets on commodity 802.11 cards.
To address both ISI and ICI, OFDM transceivers reserve the beginning portion of the symbol time as
the guard time and perform the Fourier transform only on the non-guard time portion of the symbol
time. The non-guard time portion of the symbol is often called the FFT integration time because the
Fourier transform is performed only on that portion of the symbol.
Delays shorter than the guard time do not cause ICI because they do not allow frequency
components to leak into successive symbol times. Selecting the guard time is a major task for
designers of OFDM systems. The guard time obviously reduces the overall throughput of the system
because it reduces the time during which data transmission is allowed. A guard time that is too short
does not prevent interference but does reduce throughput, and a guard time that is too long reduces
throughput unnecessarily.
Cyclic Extensions (Cyclic Prefixes)
The most straightforward method of implementing the guard time would be simply to transmit
nothing during the guard time, as shown in Figure 13-5.
Figure 13-5. Naive implementation of guard time (do not do this!)
Simplistic implementations of the guard time can destroy orthogonality in the presence of common
delay spreads. OFDM depends on having an integer number of wavelengths between each of the
carriers. When the guard time is perfectly quiet, it is easy to see how a delay can destroy this
necessary precondition, as in Figure 13-5. When the two subcarriers are added together, the spectral
analysis shows subcarrier 1 (two cycles/symbol) as a strong opdpresence and a relatively smaller
amount of subcarrier 2 (three cycles/symbol). In addition, the spectral analysis shows a large
number of high-frequency components, muddying the waters further. These components are the
consequence of suddenly turning a signal "on."
Solving the problems associated with a quiet guard time is quite simple. Each subcarrier is extended
through the FFT integration period back through the preceding guard time. Extending each subcarrier
(and hence the entire OFDM symbol) yields a Fourier transform that shows only the amplitudes of the
subcarrier frequencies. This technique is commonly called cyclic extension, and it may be referred to
as the "cyclic prefix extension." The guard time with the extended prefix is called the cyclic prefix.
In Figure 13-6, the cyclic prefix preserves the spectral analysis. Subcarrier 1 was not shifted and is
not a problem. Subcarrier 2 is delayed, but the previous symbol appears only in the guard time and
is not processed by the Fourier transform. Thanks to the cyclic prefix extension, when subcarrier 2 is
processed by the Fourier transform, it is a pure wave at three cycles per integration time.
Figure 13-6. Cyclic prefix extension
Windowing
One further enhancement helps OFDM transceivers cope with real-world effects. Transitions can be
abrupt at symbol boundaries, causing a large number of high-frequency components (noise). To
make OFDM transmitters good radio citizens, it is common to add padding bits at the beginning and
end of transmissions to allow transmitters to "ramp up" and "ramp down" from full power. Padding
bits are frequently needed when error correction coding is used. Some documentation may refer to
the padding as "training sequences."
Windowing is a technique used to bring the signal for a new symbol gradually up to full strength
while allowing the old symbol to fade away. Figure 13-7 shows a common windowing function based
on a cosine curve. At the start of the symbol period, the new function is brought up to full strength
according to the cosine function. When the symbol ends, the cosine curve is used to fade out the bits
at the end of the symbol.
Figure 13-7. Cosine windowing technique
OFDM as Applied by 802.11a
802.11a is not a radical application of OFDM. The task group responsible for standardizing OFDM took
the middle ground to apply OFDM to wireless LAN.
OFDM Parameter Choice for 802.11a
When choosing OFDM parameters, there are usually three given items of information. Bandwidth is
fixed, often by regulatory authorities. Delay is determined by the environment in which the OFDM
system will operate; most office buildings generally show a delay spread of 40-70 ns, though in some
environments, the delay spread can approach 200 ns. Finally, the bit rate is usually a design goal,
although the goal is usually "make the bit rate as high as possible, given the constraints of the other
parameters."
One common guideline is that the guard time should be two to four times the average delay spread.
As a result, the 802.11a designers selected a guard time of 800 ns. Symbol duration should be much
larger than the guard time, but within reason. Larger symbol times mean that more subcarriers can
fit within the symbol time. More subcarriers increase the signal-processing load at both the sender
and receiver, increasing the cost and complexity of the resulting device. A practical choice is to select
a symbol time at least five times the guard time; 802.11a matches the 800-ns guard time with a 4ms symbol time. Subcarrier spacing is inversely related to the FFT integration time. 802.11a has a
3.2-ms integration time and a subcarrier spacing of 0.3125 MHz (1/3.2 ms).
Operating channels in 802.11a are specified as 20 MHz-wide. The bandwidth of an operating channel
is a design decision. Wider operating channels have higher throughput, but fewer operating channels
fit into the assigned frequency spectrum. The use of a 20-MHz operating channel allows for
reasonable speeds on each channel (up to 54 Mbps), as well as a reasonable number of operating
channels in the assigned spectrum. 802.11a offers a wide variety of choices in modulation and coding
to allow for a trade-off between robust modulation and conservative coding, which yields low, reliable
throughput, finer-grained modulation, and aggressive coding, resulting in higher, yet somewhat more
fragile, throughput.
Structure of an Operating Channel
Like the DS PHYs, the OFDM physical layer organizes the spectrum into operating channels. Each 20MHz channel is composed of 52 subcarriers. 4 of the subcarriers are used as pilot carriers for
monitoring path shifts and ICI, while the other 48 subcarriers are used to transmit data. Subcarriers
are spaced 0.3125 MHz apart. As shown in Figure 13-8, channels are numbered from -26 to 26.
Subcarrier 0 is not used for signal-processing reasons.
Pilot subcarriers are assigned to subcarriers -21, -7, 7, and 21. To avoid strong spectral lines in the
Fourier transform, the pilot subcarriers transmit a fixed bit sequence specified in 802.11a using a
conservative modulation technique.
Figure 13-8. Structure of an OFDM channel
Subchannel modulation techniques
802.11a uses a technique called quadrature amplitude modulation (QAM) on each of the subcarriers
to transmit data. QAM encodes data on a single carrier wave, but that carrier wave is composed of
two components, called the in-phase and quadrature components. QAM performs amplitude
modulation on both components; that is, it varies the size of the carrier wave based on the input.
The main carrier wave is called in the in-phase component, abbreviated I. Lagging behind a quartercycle out of phase is the quadrature component, abbreviated Q. (Alternatively, QAM can be modeled
mathematically as a single wave by using complex numbers.) Essentially, the amplitude of the
composite signal and phase shifts within it both encode information.[*]
[*]
The phase shift modulation used by direct-sequence 802.11 PHYs is a special case of QAM, where the amplitude does not
change but the phase does.
QAM is widely used in radio transmission systems. In North American television transmission, the inphase component carries luminance, or brightness information, while the quadrature component
carries the color information. AM radio is capable of stereo reception as well; a QAM system system
called C-QUAM encodes the stereo information on the quadrature component.
Both television and radio are analog systems. In digital systems, the in-phase and quadrature
components are restricted to particular levels, or quantized. With both components restricted to a
particular set of levels, the result is that a constellation is formed. Constellations are usually
displayed by plotting allowed values of the in-phase signal and quadrature signal against each other
on a two-dimensional plot. Each point in the constellation is assigned to a symbol, and each symbol
is assigned to a specific bit value. QAM is described in writing with the number of bits in the
constellation, such as 64-QAM. Because the number of quantized values should always be a multiple
of two in each direction, the number preceding QAM is always a power of two. 802.11a uses square
constellations, and many of the proposals for faster codes use square constellations as well, which
means that both directions must be an even power of two, and the number is also a perfect square
(16-QAM, 64-QAM, 256-QAM).
To transmit data at higher rates, define a constellation with more points. However, as the data rate
increases, the received signal quality must be good enough to distinguish between adjacent points in
the constellation. As they are packed closer together, the acceptable error range around a point
shrinks. 802.11a specifies the maximum acceptable errors from a constellation point as part of the
PHY. Figure 13-10 shows the constellations used by 802.11a. At the lowest bit rates are BPSK and
QPSK, the two phase-shift keying modulations used by the direct sequence PHYs.
Figure 13-9. Constellations used by 802.11a
The total capacity of the radio channel is obtained by multiplying the number of subchannels by the
number of bits per channel. A radio channel using 64-QAM on each subchannel can carry six bits per
subchannel. In 802.11a, there are 48 subchannels, which gives a capacity of 288 bits per channel.
However, there is a second attribute to account for. Most radio channels are not capable of error-free
operation, and incorporate an error correction code.
Forward error correction with convolutional coding
Strictly speaking, forward error correction is not part of OFDM. However, OFDM is used in
applications for which the signal is subject to narrowband interference or frequency-specific
narrowband fading, also known as deep fading. When fading occurs, a channel's ability to carry data
may go to zero because the received amplitude is so small. To keep a few faded channels from
driving the bit error rate to the sky, OFDM implementations often apply an error correction code
across all the subchannels. Implementations that use an error correction code in conjunction with
OFDM are sometimes called coded OFDM (COFDM).
Coded OFDM uses a forward error correction (FEC) code on each channel. Forward error correction is
a system that allows a receiver to detect corrupted bits and repair the transmission, provided that
the fraction of lost or damaged bits is small enough. To provide the error-correction capability,
redundant bits are added to the data stream. FEC codes work by creating a state engine that
depends on the data bits currently being transmitted and encoding them across several symbols. By
smearing noise out through several transmitted symbols, transient noise may be reduced to the point
where resulting errors do not affect the receiver's ability to reconstruct the bit stream.
Two classes of forward error correction codes can be used. Block codes work by taking input blocks of
a fixed size, while convolutional codes work on streams of any length. Block codes are used for
applications in which the data is naturally stored, transmitted, or processed in chunks. DVDs, CDs,
and hard disks use a block code. Convolutional codes, however, are the easy choice for wireless LANs
because the frames being transmitted may be a variety of sizes. [*]
[*]
In 1993, it was discovered that two convolutional codes can be combined into a turbo code. Turbo codes are extremely
powerful, though at the cost of additional latency over convolutional codes. They are used in applications where the additional
latency is not a major factor, such as space and satellite communications. For further information about turbo codes, see the Jet
Propulsion Laboratory's site at http://www331.jpl.nasa.gov/public/JPLtcodes.html, which features overview comparing the
performance of turbo codes to the error correction codes used on the Voyager and Cassini probes. Turbo codes also underpin
1xEV-DO, Qualcomm's 3G Internet access system.
Convolutional codes have two main parameters. The constraint length determines how long a data bit
is averaged into successive transmissions. Longer constraint lengths average data bits (and hence,
any noise) over a longer time period, improving the reliability of the transmission at the cost of more
complexity in the decoder. For relatively short constraint lengths, the Viterbi algorithm is used; it is
recommended by the 802.11a standard itself. One of the reasons for the recommendation is that the
Viterbi decoder is a maximum likelihood decoder. The results from a Viterbi decoder are the most
likely data bits that were transmitted. Not all decoders are maximum likelihood, though it is
obviously an important property for digital transmission systems. 802.11a uses a constraint length of
7, which is the limit of where a Viterbi decoder can practically be used. (Complexity of Viterbi
decoding grows exponentially with constraint length.)
The second parameter, the coding rate (R), determines how many redundant bits are added. It is
expressed as the number of data bits transmitted as a ratio of the total number of coded bits. A
convolutional code with R=1/2 transmits one data bit for every two code bits. More aggressive codes
may have less redundant information, such as R=3/4, where only 25% of the bits are redundant.
Selecting a coding rate is a matter of engineering. As the code rate decreases, more code bits are
available to correct errors, and the code becomes more robust. However, the price of robustness is
decreased throughput. 802.11a specifies three code rates ranging from 1/2 at the most conservative
end to 3/4 at the most aggressive end.[*]
[*]
The basic convolutional code specified by 802.11a uses a constraint length of 7 and a code rate of 1/2. This code was used on
the Voyager deep-space probe.
Convolutional code rates may be transformed by puncturing. A punctured code changes a lower-rate
convolutional code into a higher-rate code by discarding some of the coded bits. When data is coded,
the lower-rate code is run. In the case of 802.11a, the lowest code rate is 1/2, which doubles the
size of the data. To puncture to a 3/4 rate, one third of the coded data is discarded, or "punctured."
To puncture to a 2/3 rate, 30% of the coded data is discarded. The coded data, minus the dropped
bits, are transmitted. At the receiver, dummy bits are stuffed into the decoder to stand in for the
punctured bits. Figure 13-10 illustrates the process to puncture a rate 1/2 code to a rate 3/4 code.
One of the advantages of using a punctured code is that it can be implemented as single
convolutional coder under software control, and different puncturing patterns can be used to alter the
rate in software.
Figure 13-10. Puncturing to increase the code rate
Subchannel interleaving
Each operating channel has 48 constituent subcarriers. In essence, each operating channel is a
multiplexed sum of 48 separate data streams. The incoming stream of coded bits must be mapped
on to the correct subcarrier. Rather than use a simple round-robin algorithm that maps the first bit
on to the first subcarrier, the second bit on to the second subcarrier, and so on, 802.11a uses a pair
of interleaving rules. The first rule ensures that bits in sequence are transmitted on widely separated
subcarriers, and the second ensures that bits in sequence are mapped on to different constellation
points. As an example, Figure 13-11 illustrates the interleaver in operation with 16-QAM, which has
192 bits. Each of the 192 bits must be mapped on to one of the 48 data-carrying subcarriers. The
figure plots the bit position against the subcarrier, clearly showing how each subcarrier is responsible
for four coded bits.
Figure 13-11. Interleaving with 16-QAM
Operating Channels
Channels in the 5-GHz band are numbered starting every 5 MHz according to the following formulas:
Obviously, each 20-MHz 802.11a channel occupies four channel numbers. The recommended channel
use is given in Table 13-1. 802.11a was originally designed for the United States. European
channelization was added as part of 802.11h in late 2003, and Japanese operation was added with
802.11j in late 2004.[*]
[*]
European radio regulations are standardized by the European Conference of Postal and Telecommunications Administrations
(CEPT). The European Radiocommunications Office develops regulations for CEPT; the 5 GHz radio LAN regulations are
ERC/DEC/(99)23 (available at http://www.ero.dk/doc98/Official/Pdf/DEC9923E.PDF). Japanese operation was approved by the
Ministry of Internal Communications (http://www.soumu.go.jp/), in Articles 49.20 and 49.21
(http://www.soumu.go.jp/joho_tsusin/eng/Resources/Legislation/MRA/020226_00.html).
Table 13-1. Operating channels for 802.11a/j
Frequency
Allowed in?
Allowed powera
4.920-4.980
GHz
Japan
250 mW EIRP and
< 1 W, or 10 mW
EIRP
5.040-5.080
GHz
Japan
5.15-5.25 GHz Japan
250 mW EIRP and
< 1 W, or 10 mW
EIRP
200 mW (< 10
mW/MHz)
Channel
numbers
Center
frequency
(GHz)
240
4.920
244
4.940
248
4.960
252
4.980
8
5.040
12
5.060
16
5.080
34
5.170
Frequency
U-NII lower
band
Allowed in?
United States
Channel
numbers
Center
frequency
(GHz)
38
5.190
42
5.210
46
5.230
36
5.180
40
5.200
44
5.220
48
5.240
52
5.260
56
5.280
60
5.300
64
5.320
100
5.500
104
5.520
108
5.540
112
5.560
116
5.580
120
5.600
124
5.620
128
5.640
132
5.660
136
5.680
140
5.700
800 mW
149
5.745
(50 mW/MHz)
153
5.765
157
5.785
161
5.805
a
Allowed
mW/MHz)power
40 mW (2.5
mW/MHz)
(5.15-5.25
GHz)
U-NII midband
United States
200 mW (12.5
mW/MHz)
(5.25-5.35
GHz)
5.470-5.725
U-NII upper
band
(5.725-5.825
GHz)
a
b
Europe (CEPT) (allowed in
United States as of 11/2003,
subject to U-NII rulesb)
1 W EIRP
In the United States, the allowed power is the maximum output power using a 6-dBi antenna gain.
The additional spectrum was opened in FCC 03-287
(http://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC-03-287A1.pdf), but test procedures were
still in development as this book went to press. Without testing, no devices were able to be certified
to operate in this spectrum in the United States.
There is one other feature of note about the operating bands. As with the DS PHYs, a transmit mask
limits power leakage into the side bands. The mask is shown in Figure 13-12.
Figure 13-12. Transmit spectrum mask for 802.11a
Figure 13-13 gives an overall view of the 802.11a channels available in the U.S. Four channels are
available in each of the U-NII bands in the U.S. In the two lower U-NII bands, channels are allowed
to overlap, and a 30-MHz guard band is present at both the lower end of the low U-NII band and the
upper end of the mid U-NII band.
Figure 13-13. Operating bands from Table 13-1
OFDM PLCP
Like all the other physical layers, the OFDM PHY includes its own PLCP, which adds physical layerspecific framing parameters.
Framing
The OFDM PHY adds a preamble and a PLCP header. It also adds trailing bits to assist the encoding
schemes used. This section divides the PLCP frame logically, but some components span different
fields in the protocol unit. Figure 13-14 is the jumping-off point for discussion of the OFDM frame.
Figure 13-14. OFDM PLCP framing format
Figure 13-15 shows the start of a frame, but includes the guard intervals and windowing used by the
transmitter. The preamble lasts 16 ms, which is evenly divided between short and long training
sequences; the difference between the two is described in the next section. After the preamble, one
OFDM symbol carries the Signal field, then a variable number of data symbols carry the end of the
PLCP header, the MAC payload, and the trailer. All symbols use a modified cosine window to ensure
smooth transitions. After the short preamble, which is used to synchronize frequencies, a guard time
protects against multipath fading.
Figure 13-15. Preamble and frame start
Preamble
As with all other common IEEE 802 networks, and certainly all 802.11 physical layers, the OFDM
physical protocol unit begins with a preamble. It is composed of 12 OFDM symbols that synchronize
various timers between the transmitter and the receiver. The first 10 symbols are a short training
sequence, which the receiver uses to lock on to the signal, select an appropriate antenna if the
receiver is using multiple antennas, and synchronize the large-scale timing relationships required to
begin decoding the following symbols. The short training sequences are transmitted without a guard
period. Two long training sequences follow the short training sequences. Long training sequences
fine-tune the timing acquisition and are protected by a guard interval.
Header
The PLCP header is transmitted in the Signal field of the physical protocol unit; it incorporates the
Service field from the Data field of the physical protocol unit. As shown in Figure 13-16, the Signal
field incorporates the Rate field, a Length field, and a Tail field.
Figure 13-16. Signal field of OFDM PLCP frame
Rate (4 bits)
Four bits encode the data rate. Table 13-2 shows the bits used to encode each of the data
rates. See the section "OFDM PMD" later in this chapter for details on the encoding and
modulation scheme used for each data rate.
Table 13-2. Rate bits
Data rate (Mbps)
Bits (transmission order)
6
1101
9
1111
12
0101
18
0111
24
1001
36
1011
48
0001
54
0011
Length (12 bits)
12 bits encode the number of bytes in the embedded MAC frame. Like most fields, it is
transmitted least-significant bit to most-significant bit. The length is processed by a
convolutional code to protect against bit errors.
Parity (1 bit) and Reserved (1 bit)
Bit 4 is reserved for future use and must be set to 0. Bit 17 is an even parity bit for the first 16
Signal bits to protect against data corruption.
Tail (6 bits)
The Signal field ends with six 0 tail bits used to unwind the convolutional code. As such, they
must by definition be processed by the convolutional code.
Service (16 bits)
The final field in the PLCP header is the 16-bit Service field. Unlike the other components of the
PLCP header, it is transmitted in the Data field of the physical protocol unit at the data rate of
the embedded MAC frame. The first eight bits are set to 0. As with the other physical layers,
MAC frames are scrambled before transmission; the first six bits are set to 0 to initialize the
scrambler. The remaining nine bits are reserved and must set to 0 until they are adopted for
future use.
Data
The encoding scheme used for the data depends on the data rate. Before transmission, data is
scrambled, as it is with the other physical layers. The Service field of the header is included in the
Data field of the physical protocol unit because it initializes the scrambler.
Trailer
The Data field of the physical protocol unit ends with a trailer. (The 802.11a specification does not
call the ending fields a trailer, but it is a descriptive term.) It is composed of two fields:
Tail (6 bits)
Like the tail bits in the PLCP header, the tail bits appended to the end of the MAC frame bring
the convolutional code smoothly to an end. Six bits are required because the convolutional
code has a constraint length of seven.
Pad (variable)
As used by 802.11a, OFDM requires that fixed-size blocks of data bits be transferred. The Data
field is padded so that its length is an integer multiple of the block size. The block size depends
on the modulation and coding used by the data rate; it is discussed in the next section.
OFDM PMD
The OFDM PHY uses a cocktail of different modulation schemes to achieve data rates ranging from 6
Mbps to 54 Mbps. In all cases, the physical layer uses a symbol rate of 250,000 symbols per second
across 48 subchannels; the number of data bits per symbol varies. An OFDM symbol spans all 48
subchannels.
Encoding and Modulation
There are four rate tiers with the OFDM PHY: 6 and 9 Mbps, 12 and 18 Mbps, 24 and 36 Mbps, and
48 and 54 Mbps. Support is required for 6, 12, and 24 Mbps, which are lowest speeds in each of the
first three tiers, and therefore the most robust in the presence of interference. The lowest tier uses
binary phase shift keying (BPSK) to encode 1 bit per subchannel, or 48 bits per symbol. The
convolutional coding means that either half or one quarter of the bits are redundant bits used for
error correction, so there are only 24 or 36 data bits per symbol. The next tier uses quadrature
phase shift keying (QPSK) to encode 2 bits per subchannel, for a total of 96 bits per symbol. After
subtracting overhead from the convolutional code, the receiver is left with 48 or 72 data bits. The
third and fourth tiers use generalized forms of BPSK and QPSK known as quadrature amplitude
modulation (QAM). 16-QAM encodes 4 bits using 16 symbols, and 64-QAM encodes 6 bits using 64
symbols. The third tier uses 16-QAM along with the standard R=1/2 and R=3/4 convolutional codes.
To achieve higher rates with 64-QAM, however, the convolutional codes use R=2/3 and R=3/4. Table
13-3 summarizes the coding methods used by each data rate in the OFDM PHY.
Table 13-3. Encoding details for different OFDM data rates
Speed
(Mbps)
Modulation and coding
rate (R)
Coded bits per
carriera
Coded bits per
symbol
Data bits per
symbolb
6
BPSK, R=1/2
1
48
24
9
BPSK, R=3/4
1
48
36
12
QPSK, R=1/2
2
96
48
18
QPSK, R=3/4
2
96
72
24
16-QAM, R=1/2
4
192
96
36
16-QAM, R=3/4
4
192
144
48
64-QAM, R=2/3
6
288
192
54
64-QAM, R=3/4
6
288
216
72 c
64-QAM
6
288
288
a
Coded bits per subchannel is a function of the modulation (BPSK, QPSK, 16-QAM, or 64-QAM).
b
The data bits per symbol is a function of the rate of the convolutional code.
Radio Performance: Sensitivity and Channel Rejection
Speed
(Mbps)
Modulation and coding
rate (R)
Coded bits per
carriera
Coded bits per
symbol
Data bits per
symbolb
c
Although no rate has been standardized without a convolutional code, many products offer a mode
where it is dropped for additional throughput.
Radio Performance: Sensitivity and Channel Rejection
Like other physical layers, 802.11a specifies minimum performance requirements for the receiver,
which are shown in Table 13-4. Minimum sensitivity has been discussed for the other physical layers.
The only feature of note in 802.11a is that the wide range in data speeds means that a minimum
performance requirement is quoted for each data speed. In comparison with the requirements laid
down by the direct-sequence layers, 802.11a is just as stringent. 802.11a requires a -76 dBm
sensitivity, which is comparable to the 18 Mbps and 24 Mbps data rates in 802.11a.[*]
[*]
Note, however, that path loss is worse at the 802.11a frequencies.
More interesting is the specification of channel rejection. As with the other physical layers, begin by
injecting a signal slightly (3 dB) above the minimum sensitivity on a given channel. On either
rejection test, bring up a second signal on either an adjacent or nonadjacent channel. When the
channel under test suffers a 10% frame error rate, note the difference in power between the two
channels.
What the table says is that as the data rate increases, the more easily a signal is disrupted at the
receiver. If an 802.11a network is built too dense, so that 54 Mbps signals from adjacent APs are
regularly received in the middle of the two, it is possible that the client radio chipsets will not be able
to decode the transmissions. However, this scenario is unlikely due to the relatively short range of 54
Mbps transmissions. Furthermore, many chipsets will perform better than the standard requires, but
most vendors do not quote rejection in the data sheets for their cards.
Table 13-4. Receiver performance requirements
Data rate
(Mbps)
Minimum sensitivity
(dBm)
Adjacent channel
rejection (dB)
Nonadjacent channel
rejection (dB)
6
-82
16
32
9
-81
15
31
12
-79
13
29
18
-77
11
27
24
-74
8
24
36
-70
4
20
48
-66
0
16
54
-65
-1
15
Clear Channel Assessment
The OFDM PHY specification leaves implementers a great deal of latitude in selecting techniques for
noting a busy channel. Received signal strength thresholds determine whether the channel is in use,
but the main guideline for 802.11a equipment is that it must meet certain performance standards.
Implementations are free to use the Packet Length field from the PLCP header to augment clear
channel assessment, but this is not required.
Transmission and Reception
The block diagram for an 802.11a receiver is shown in Figure 13-17. When a frame is ready for
transmission, the 802.11a interface runs the following procedure:
1. Select a transmission rate. Algorithms for selecting a data rate are implementation-dependent,
so products sold by different vendors may select different rates under identical circumstances.
The rate dictates the modulation and convolutional code, as well as the number of data bits per
subcarrier. See Table 13-3.
2. Transmit the PLCP preamble, which consists of long and short training sequences.
3. Begin transmission of the PLCP header with the SIGNAL field, which is not scrambled. It is
coded with the convolutional coder.
4. Create the data field of the packet.
a. Create the SERVICE field, which is currently set to all zeros. It has seven zeros for
scrambler initialization, and nine reserved bits set to zero.
b. Append the data.
c. Put on the six zero tail bits.
d. Pad with zeros to a multiple of the data bit per subcarrier block size.
5. Scramble the data, which helps avoid long strings of zeros or ones.
6. Encode the data with the convolutional coder. If necessary, puncture the output of the
convolutional code to generate an encoded string at a higher rate than 1/2.
7. Divide the coded data into blocks for processing. The block size depends on the modulation rate
for the data symbols.
a. Perform the interleaving process and map bits from the block on to the 48 subcarriers.
b. Insert the four pilot subcarriers at the designated locations.in the channel.
c. Use the inverse Fourier transform to convert the frequency-domain data into time-domain
data for transmission.
8. Repeat step 5 for each data block until there are no more.
8.
Figure 13-17. A transceiver block diagram
Acknowledgment
Support of the 6, 12, and 24 Mbps data rates is required by 802.11a. Upon receipt of a frame, the
802.11 MAC requires an acknowledgment. Acknowledgments must be sent at a supported data rate
for all associated stations. Most devices send acknowledgments at 24 Mbps because it minimizes the
overhead while obeying the stricture to transmit at a mandatory rate.
An example of OFDM encoding
OFDM encoding, as you can no doubt see by now, is an intense, multistep process. One of the
additions that 802.11a made to the original specification was Annex G, an encoding of Schiller's Ode
to Joy for transmission over an 802.11a network. Shortly after 802.11a was published, the IEEE
802.11 working group discovered several errors in the example and published a correction. If you are
interested in learning about OFDM encoding in detail, you can refer to this example.
Characteristics of the OFDM PHY
Parameters specific to the OFDM PHY are listed in Table 13-5. Like the physical layers presented in
Chapters 11 and 12, the OFDM PHY also incorporates a number of parameters to adjust for the delay
in various processing stages in the electronics. As a final note, the extra radio bandwidth provided
offers a great deal of throughput. In the United States, there are currently 12 channels, with the
additional space for eight channels recently cleared. 20 co-located channels operating with 50% of
the transmission time devoted to user traffic can move over 500 Mbps. This number is far greater
than any other physical layer yet standardized.
Table 13-5. OFDM PHY parameters
Parameter
Value
Maximum MAC frame
length
4,095 bytes
Slot time
9 ms
SIFS time
16 ms
Contention window
size
15 to 1,023
slots
Preamble duration
20 ms
PLCP header duration
4 ms
Receiver sensitivity
-65 to -82
dBm
Notes
The SIFS is used to derive the value of the other interframe
spaces (DIFS, PIFS, and EIFS).
Depends on speed of data transmission.
Like the other physical layers, the OFDM PHY has a number of attributes that can be adjusted by a
vendor to balance delays in various parts of the system. It includes variables for the latency through
the MAC, the PLCP, and the transceiver, as well as variables to account for variations in the
transceiver electronics.
Chapter 14. 802.11g: The Extended-Rate
PHY (ERP)
When wireless LANs first entered the mainstream computing consciousness, there was one practical
choice. 802.11b had recently been standardized, and offered the prospect of near-Ethernet speed,
which, to be fair, by that point was not very fast at all. As 802.11a emerged from the laboratory into
commercially-available chipsets, users had a desire to obtain higher speeds than 802.11b, while
retaining backwards compatibility with the installed base of 802.11b hardware. The result is 802.11g,
which offers a headline bit rate comparable to 802.11a while still operating in the microwave band.
By working at slightly less than half the frequency, 802.11g devices have better range than the 5
GHz 802.11a devices.
802.11g is not a revolutionary specification. In fact, if you read the new clauses added by the
specification, it is clear that it uses much of the existing work done elsewhere. The new physical
layers it specifies are built on existing work, and there are only slight modifications. Most of 802.11g
is occupied with providing backwards compatibility.
802.11g Components
802.11g is really several physical layer specifications in one. It adds one umbrella clause for the
Extended Rate PHY (ERP). However, there are several different "flavors" of ERP:
ERP-DSSS and ERP-CCK
These modes are backwards compatible with the original direct sequence specification (1 Mbps
and 2 Mbps) described in Chapter 11, as well as the 802.11b enhancements (5.5 Mbps and 11
Mbps) described in Chapter 12. To retain backwards compatibility, a few minor changes are
required.
ERP-OFDM
This is the major mode of 802.11g. It is essentially running 802.11a in the ISM frequency band
(2.4 GHz), with a few minor changes to provide backwards compatibility. It supports the same
speeds as 802.11a: 6, 9, 12, 18, 24, 36, 48, and 54 Mbps. Speeds of 6, 12, and 24 Mbps are
mandatory.
ERP-PBCC
This is an optional extension to the PBCC standard provided in 802.11b, and provides data
rates of 22 Mbps and 33 Mbps. Although it is part of the standard, it is not implemented by
most major chipsets in the market, and is not widely used.
DSSS-OFDM
This a hybrid scheme, which encodes packet data using the DSSS headers, and OFDM encoding
of the payload. Part of the reason for developing this implementation was for backwards
compatibility. Although the body is OFDM-modulated and unintelligible to 802.11b, information
in the headers is able to provide information on the duration of the packet. It is optional, and
not widely implemented.
Any device that implements 802.11g is required to support a few mandatory modes. For backwards
compatibility, 802.11g devices must support DSSS modulation (802.11) at 1 and 2 Mbps, and CCK
modulation (802.11b) at 5.5 and 11 Mbps. Basic OFDM support is required, and all 802.11g stations
are further required to support OFDM modulation at 6, 12, and 24 Mbps.
Compatibility Changes
The mandatory modes in 802.11g are slight modifications of existing physical layers, with a few
minor alterations made for backwards compatibility. The modifications required of 802.11g stations
assist coexistence with older implementations. They are not changes to the existing specifications. An
802.11b card will work as it always has. The only difference is that an 802.11g card will have a few
features not present in 802.11b.
802.11b devices implement two different specifications: the original, slow direct sequence (DSSS)
from the initial 802.11 standard, and the high-rate complementary code keying (CCK) PHY from
802.11b. 802.11g adopts both of those standards, and makes only a few minor changes. Naturally,
any 802.11g station must be able to hear not only older stations but other 802.11g stations, so any
802.11g-compliant station must support all the preambles and synchronization found in 802.11g.
More significantly, 802.11g stations must support the short preamble because it helps a great deal in
maintaining high throughput. 802.11g radios are required to be more sensitive to signals as well.
802.11g devices must also implement ERP-OFDM, which is based heavily on 802.11a. In fact, it looks
almost exactly like 802.11a, with a few obvious changes. Most notably, 802.11g adopts the
frequency plan of 802.11b, so there are still only three ostensibly nonoverlapping channels for use.
(See Chapter 13 for the adjacent-channel overlap map; the same note about three mostly
nonoverlapping channels applies to 802.11g as well.) It also uses interframe spacing and slot times
that are compatible with older ISM-band 802.11 stations. Regulatory structures may be different
around 802.11g and 802.11b, however. Japan allows 802.11b operation in channels 1-14, but
802.11g is only allowed on channels 1-13.
Protection
One of the major differences between 802.11b and 802.11g is protection, which is required because
of an asymmetry between the chipsets that implement the specifications. One of the problems faced
by the designers of 802.11g is that it uses a different modulation scheme than 802.11b. 802.11g
chips are built with backwards compatibility, and have no problem receiving and decoding an
802.11b signal. As is normally the case in technology, the converse is not true. 802.11b chipsets
have no way of making sense of the higher-speed 802.11g transmissions. Part of the solution is to
require that 802.11g stations transmit at a rate supported by all stations in basic service set. If an AP
is to serve both 802.11b and 802.11g stations, it will need to send Beacon frames at a frame data
rate of no higher than 11 Mbps.
The second part of the solution is to avoid interference between 802.11g and 802.11b networks. To
ensure that 802.11b stations are aware of 802.11g transmissions, 802.11g specifies a protection
mechanism to "protect" 802.11b stations from interference. The basic operation of the protection
mechanism is shown in Figure 14-1. To avoid interference during the transmission of the OFDM
frame and its acknowledgment, a slower frame is sent to update the NAV. There are two main
protection modes. More commonly, 802.11g stations will use CTS-to-self protection, as shown in
Figure 14-1 (a). When a station has a frame for transmission that needs protection, it will transmit a
CTS frame with a receiver address of its own MAC address; that is, the destination of the CTS frame
is the station itself. In the CTS, it will update the Network Allocation Vector (NAV) to tell other
stations using the physical medium that it will be using the radio link for the time necessary to
transmit the CTS, the OFDM-modulated frame, and an OFDM-modulated acknowledgment. Although
the station sends the CTS to itself, all stations on the network are required to listen to CTS frames
and update the NAV accordingly. The CTS frame is sent at the maximum speed it can be, using a
modulation that can be received by all stations. Throughput may suffer a significant hit as a result of
this exchange. A maximum Ethernet-size data frame and its acknowledgment require 294
microseconds at the highest speed, but the CTS to clear out the network requires at least 107
microseconds, and possibly more than 200 microseconds if long preambles are in use.
Figure 14-1 (b) shows the second mechanism, which is a full RTS/CTS exchange. Full RTS/CTS
exchanges are more robust against hidden nodes, but come with a cost in network capacity. As in the
CTS-to-self case, the compatibility frames used to reserve the medium may be have a transmission
time comparable to or longer than the data. Both the RTS and CTS frames will take 100
microseconds, while a maximum-size data frame and its ACK only require 300 microseconds. Based
on calculations of the time required for the extra RTS, I estimate that the use of full RTS/CTS
exchanges versus CTS-to-self to be a reduction of approximately one third in overall data
throughput. [*]
[*]
See "When is 54 Not Equal to 54: An Analysis of 802.11g Throughput" at
http://www.oreillynet.com/pub/a/wireless/2003/08/08/wireless_throughput.html.
Figure 14-1. Basic overview of protection mechanism
To ensure that the protection frames are received and processed by all stations on the network,
protection frames are transmitted using 802.11b rules. They may be transmitted using phase shift
keying at 1 Mbps or 2 Mbps, or CCK at 5.5 Mbps or 11 Mbps. Any 802.11b station will be able to
understand these modulations, and can update its virtual carrier sense accordingly. Figure 14-1
shows this by labeling the transmission modulation, and only showing the compatible modulations as
received by an 802.11b station.
Only the protection frames are required to be transmitted at the 802.11bcompatible data rates. Protection does not require 802.11g stations to use a
slower data rate for the payload data, as is commonly asserted.
Protection is activated whenever there is a need to ensure 802.11g stations do not interfere with
802.11b stations. One obvious case in which this happens is when an 802.11b station associates with
an AP. All the 802.11g stations associated with the AP will then start using protection to ensure that
the 802.11b station does not suffer from or cause interference. Protection is also activated in the less
obvious case of a co-channel AP with 802.11b-only traffic. Because the two co-channel APs share the
same physical medium, an 802.11b station using the same channel on a different AP also triggers
protection.
Protection is controlled through the ERP information element in Beacon frames. 802.11g adds a Use
Protection bit in to an information element in the Beacon. When the bit is set, stations must use
protection. When a non-802.11g-capable station associates to a wireless LAN, the protection bit will
be set. The station responsible for sending the Beacon is also responsible for deciding whether to
activate protection. In infrastructure networks, the protection activation is handled by the access
points; in independent networks, it is the Beacon generator. Protection is activated when a non802.11g station associates with the network, as well as when non-802.11g-capable stations are
transmitting in the area. Non-802.11g-capable stations may be deduced by the reception of a
management frames, including Beacons and Probe Responses, from an overlapping network that
does not indicate the data rates supported by 802.11g.
Beacons in 802.11g networks can also control the preamble length for protection purposes. In the
ERP information element, the Barker Preamble Mode bit can be used to tell associated stations
whether to use the long preamble or short preamble. If all the stations associated to a network are
capable of short preambles, the Barker Preamble Mode bit will be set to zero, and all stations will use
short preambles for efficiency. Once a station that is not capable of short preambles associates with
the network, however, the bit will be set to one, and all protection frames will use long preambles.
As with most other technology that is required to be backwards compatible, 802.11g can suffer for it.
If protection is not enabled, 802.11a and 802.11g throughput is identical. Once protection is
activated, however, my calculations showed it to be very roughly 50%, depending on the ratio of TCP
data segments to TCP ACKs. Transmitting a full-size 1,500 byte Ethernet frame over 802.11 requires
428 microseconds for both the frame and the 802.11 acknowledgment. When CTS-to-self protection
is used, the transmission time jumps to 557 microseconds; RTS/CTS adds further bloat, and requires
774 microseconds for the same transmission. Avoiding the need for protection is one of the reasons
that many 802.11g access points offer the ability to accept associations from 802.11g-capable
stations only. By shunning stations that are not capable of 802.11g speeds, there is an increased
probability that protection can remain disabled. Of course, a station that is shunned may very well
find a different network that can accommodate it on the same channel, which would still trigger the
activation of protection.
Although 802.11a and 802.11g are capable of the same speed, the extra
transmission time required for protection may cut payload data throughput by
half.
Protection is not required for the ERP-PBCC and DSSS-OFDM physical layers. As shown in Figure 142, both of them begin with an 802.11b-compatible header, and therefore the virtual carrier sense and
NAV are updated without needing to send extra frames. An 802.11b station will receive the header
and update the virtual carrier sensing mechanism based on the contents of the header. No
transmissions will be allowed until the medium lock expires. Although the 802.11b station cannot
detect the body of the frame, it is prevented from interference by the header. There is a cost to using
ERP-PBCC or DSSS-OFDM. In a sense, they are always using protection because they use a much
slower header. Both ERP-PBCC and DSSS-OFDM have, at minimum, 96 microsecond headers that can
never be dropped. In contrast, the common ERP-OFDM mode must use the same header on a CTS
frame that takes 107 microseconds to transmit, plus a 10 microsecond SIFS gap. However, ERPOFDM needs to pay the penalty only when protection is required, while the optional modes always
need to suffer the hit.
Figure 14-2. ERP-PBCC and DSSS-OFDM frame format
ERP Physical Layer Convergence (PLCP)
The ERP PHY's convergence layer is quite complicated. Due to the plethora of operational modes,
there are a number of different ways that frames will be bundled up for transmission by the radio
interface. Each of the modes described at the start of the chapter has its own physical layer framing.
ERP-OFDM Framing
This mode is the meat of the specification. All stations are required to implement it, and it is what is
colloquially meant by "802.11g support" because most stations default to using it. In practice,
understanding ERP-OFDM physical framing will provide you with the understanding of most 802.11g
transmissions in the air.
The format of the ERP-OFDM frame at the physical layer, shown in Figure 14-3, is nearly identical to
802.11a. ERP-OFDM uses an identical logical protocol data unit, shown as the top line of the figure.
In fact, the only major difference from 802.11a is that the frame is followed by a six microsecond idle
time called the signal extension, which is shown in the second line of the figure, where the logical
construction is assembled for transmission. The reason for the extra 6 ms is to make timing
calculations and frame rates identical to 802.11a. 802.11a uses a 16 ms SIFS time, which is used in
part to finish decoding the previous frame. For backwards compatibility, 802.11g uses the 10 ms idle
time used by 802.11b. To provide enough time for the decoding process to finish, 802.11g adds 6 ms
idle time to the end of the frame, leaving a 16 ms gap for the hardware to run its decode process. Of
course, the NAV is set so that the virtual carrier sense mechanism reports the medium idle after the
signal extension period completes. In the lower line, the OFDM transmissions are shown. It is
identical to 802.11a.
Single-Carrier Framing with 802.11g
Although the OFDM modulation discussed in the previous section is by far the most common, it is
also possible to use 802.11b-compatible framing directly around the higher speed body. Traditional
framing is used with both the packet binary convolution coding and the DSSS-OFDM layer. Because
older 802.11b stations can read the frame header, they can avoid using the medium during
transmission and do not require protection. Figure 14-4 shows the use of traditional single-carrier
framing in 802.11g.
Preamble
The Preamble is identical to the 802.11b preamble discussed in Chapter 12, and consists of a
synchronization field followed by a start of frame delimiter. It may be either the long preamble
of 144 bits or the short preamble of 72 bits. In either case, the Preamble is transmitted at 1
Mbps using DBPSK modulation. Before modulation, the data is scrambled just as it is in
802.11b.
Figure 14-3. ERP-OFDM PLCP frame format
Figure 14-4. Long preamble ERP PLCP frame format
PLCP Header
The PLCP header is identical to the PLCP header discussed in Chapter 12. It consists of the
Signal field, Service field, a Length field, and a PLCP-layer CRC check. The Length and CRC
fields are identical to the 802.11b interpretation.
Signal field
This field is used to show the rate at which the PLCP payload (the MAC frame) is modulated.
Initially, it was defined as the multiplier of 100 kbps that would result in the encoding rate.
With the field defined as only eight bits, it would limit the encoding speed to 25.5 Mbps. To
accommodate increased speeds, the Signal field consists of a label that describes the speed, as
shown in Table 14-1. The signal field is not needed to tell the receiver what the encoding rate
of a DSSS-OFDM frame is because there is a separate OFDM header to perform that task.
Table 14-1. Value of the SIGNAL field in the single-carrier frame
header
Speed
Signal field value
(hex/binary)
Signal field value,
decimal
1 Mbps (ERP-DSSS)
0x0A (0000 1010)
10
2 Mbps (ERP-DSSS)
0x14 (0001 0100)
20
5.5 Mbps (ERP-CCK, ERPPBCC)
0x37 (0011 0111)
55
11 Mbps (ERP-CCK, ERPPBCC)
0x6E (0110 1110)
110
22 Mbps (ERP-PBCC)
0xDC (1101 1100)
220
33 Mbps (ERP-PBCC)
0x21 (0010 0001)
33
Any DSSS-OFDM speed
0x1E (0001 1110)
30
Service field
The service field, shown in Figure 14-5, contains control bits to help the receiver decode the
frame. As discussed previously, bits 0, 1, and 4 are reserved and must be set to zero. In all
802.11g stations, the transmit and symbol clocks are locked, so bit 2 is always set to 1. Bit 3 is
set when the frame body is modulated with PBCC, and set to zero for DSSS, CCK, and DSSSOFDM modulations. The last three length extension bits are used to assist receivers in
determining the frame length in bytes from the Length field, which is expressed in terms of the
number of microseconds required for transmission. The standards have a complicated set of
rules for when the bits must be set, but they are beyond the level of detail required for this
book, especially since the single-carrier framing technique is uncommon.
Figure 14-5. 802.11g SERVICE field
Frame Body
The final component of the PLCP frame is its payload, which is the MAC frame, modulated
either by PBCC or OFDM. Details of the frame body modulation will be discussed in subsequent
sections.
PBCC coding
To transmit a frame using PBCC, the frame data is first run through a convolutional code. The frame
is broken into 2-bit elements, and are used as input into a convolutional code that outputs three bits.
Each 3-bit block is mapped on to a symbol using 8PSK. Receivers will reverse the process, translating
a single phase shift into one of eight symbols. After translating the symbol to a three-bit sequence,
the convolutional code is used to remove the redundant bit and recover the original data. See Figure
14-6.
Figure 14-6. PBCC processing
Achieving a 22 Mbps data rate with this physical encoding is straightforward. The symbol clock
continues to run at 11 MHz, just as in 802.11b, but each symbol is now capable of transmitting two
bits. To run at 33 Mbps, the symbol rate for the data portion of the frame must be increased to 16.5
MHz. At two bits per symbol, the overall data rate is 33 Mbps. Clock speed switching must occur
between the PLCP preamble and its payload.
DSSS-OFDM framing
DSSS-OFDM is a hybrid framing technique. The higher-layer packet is encoded with OFDM, and the
OFDM-modulated packet is framed with a traditional-single carrier header, as shown in Figure 14-7.
Headers are transmitted in accord with 802.11b (including the data scrambler used by 802.11b), but
the frame body is modulated using OFDM, in a process identical to the encoding used in 802.11 and
described in Chapter 13. Although similar to the encoding in 802.11a, the DSSS-OFDM framing
eliminates the initial short training sequences. It also adds a 6 ms signal extension field to allow the
convolution decode extra time to finish. The transition from direct sequence modulation to OFDM that
occurs at the end of the PLCP header is somewhat complex from a radio engineering point of view,
but is beyond the scope of this book.
ERP Physical Medium Dependent (PMD) Layer
Once the PLCP frame is ready for transmission, it is dispatched to the Physical Medium Dependent
(PMD) layer. The PMD is responsible for taking the data and sending it out the antenna. Due to the
wide variety of modulation schemes that might be used, an 802.11g transceiver must implement
several different transmission modes, either wholly or partially, and switch between them as needed.
Some functions, however, are shared by all transceivers regardless of operating mode.
Sidebar 3.1. Atheros "Super G" Extensions
Although the speed has increased dramatically in the past several years, wireless
networks are still not fast. Under optimum conditions, a standards-based wireless
network can achieve about 30 Mbps. In the race to get the highest number possible on
the side of the box, nearly all chipset manufacturers have implemented extended
functions to increase speed. By far, the most notable (or notorious) was Atheros' Super
G enhancements, with three good features, and one that has drawn a great deal of fire.
Block acknowledgment (also called frame bursting)
In the standard 802.11 radio link management scheme, every data frame must be
followed by an acknowledgment, after which the station must re-contend for the
medium. Block acknowledgments allow a station to transmit several frames,
separated only by the short interframe space so there is no contention for the
medium. The sequence is then followed by a single block acknowledgment. Block
acknowledgment developed according to the current 802.11e drafts is generally
interoperable between vendors.
Packet clustering
The 802.11 frame can hold much more data than a standard 1,500 byte Ethernetsize frame. Using the full payload size improves the payload-to-overhead ratio,
and increases speed.
Hardware data compression
By compressing data prior to transmission, it requires less time to transmit and
gives an effective throughput increase to the link. Hardware compression is most
effective when there is redundant data, such as uncompressed text. It is not as
effective when transmitting compressed images.
Channel bonding (also called "turbo mode")
Instead of using a single 22 MHz channel, the bonding feature doubles the
bandwidth. With more spectrum to spread the signal over, a higher number of bits
can be pushed through the now-wider channel. To ensure that the wider bandwidth
stays within regulatory limits, channel bonding is restricted in some products to
operate in the middle of the band on channel 6.
By far the most controversial feature of the Super G feature set is the channel bonding.
In late 2003, Broadcom accused Super G of interfering with regular 802.11g
transmissions, and had a private demonstration at Comdex. (I was barred from
observing.) Later tests revealed some interference, but found that Broadcom's chipsets
were more susceptible to degredation than those of other vendors. Atheros has
implemented a mode which monitors for standard 802.11g transmissions, and refrains
from using channel bonding if other networks are present.
The main lesson is that any channel bonding feature is likely to increase interference by
sucking up more of the available scarce spectrum. Spectrum-hungry transmissions
modes are unsuitable for use in a large area requiring multiple access points because
they cut the number of available channels from an already-too-low three to one. Deploy
a channel bonding device in your home, but keep it off your company's network.
Figure 14-7. DSSS-OFDM frame format
Clear Channel Assessment (CCA)
Only one CCA mode is defined for 802.11g, which combines a minimum energy threshold with the
ability to decode a signal. Energy detection is based on receiving a valid signal at the start of a
transmission slot with a signal power of -76 dBm or greater. As a performance requirement, within a
specified window, the PHY should have a high probability of correctly reporting the medium busy.
Both the time window and the probability are shown in Table 14-2.
Table 14-2. 802.11g CCA performance requirements
Long slot (20 ms)
Short slot (9 ms)
CCA time
15 ms
4 ms
Detection probability
>99%
>90%
CCA is integrated with a PLCP-level virtual carrier sense. When a PLCP header is received, it will
include a length field that indicates the amount of time the medium will be busy. The physical layer
will continue to report the medium busy for that time period, even if the physical signal is lost. (Note
that this is similar in concept and operation to the Network Allocation Vector at the MAC layer.) Part
of the reason for doing this is that not all implementations will support all the transmission modes, so
it is important that the physical layer correctly avoids interfering with transmissions it cannot
demodulate.
Reception Procedure
802.11g stations have a more complicated procedure for receiving frames than chips that implement
other standards because of the choice and backwards compatibility. When an incoming frame is
detected, an 802.11g station will need to detect it and then demodulate it with the correct physical
layer.
1. Is the preamble an OFDM preamble (like 802.11a), or the traditional single-carrier preamble
used by 802.11b? Frames modulated with OFDM will be processed exactly the same as 802.11a
frames, just received on a different frequency.
2. If the frame is not an OFDM frame, it must decode the preamble, and find the end of the
preamble to see the PCLP SIGNAL and SERVICE fields.
3. By decoding the PLCP header, the appropriate modulation can be employed to demodulate the
frame body.
a. If the SERVICE field indicates that the frame is modulated with PBCC, the PBCC reception
procedures will be triggered.
b. If the SERVICE field doesn't indicate the presence of PBCC, the data rate is checked. Data
rates of 1 and 2 Mbps are processed with the DSSS reception algorithm, and data rates of
5.5 Mbps and 11 Mbps are processed according to the CCK reception algorithm; both are
described in Chapter 12.
c. If the SIGNAL field indicates a speed of 3 Mbps, DSSS-OFDM reception is used. It will
switch the demodulator to receiving OFDM at the end of the PCLP header.
Characteristics of the ERP PHY
802.11g has very similar characteristics to 802.11a, with one notable exception. Although each
channel has similar performance to an 802.11a channel, there are only three channels. If each
channel is run at the highest data rate and 50% efficiency, the total aggregate throughput is only 81
Mbps. It is a high number, but does not begin to approach 802.11a.
Table 14-3. ERP PHY parameters
Parameter
Value
Maximum MAC
frame length
4,095 bytes
Slot time
20 ms
9 ms
Notes
If the network consists only of 802.11g stations, the slot time may
be shortened from the 802.11b-compatible value to the shorter
value used in 802.11a.
SIFS time
10 ms
The SIFS is used to derive the value of the other interframe spaces
(DIFS, PIFS, and EIFS).
Signal extension
time
6 ms
Every 802.11g packet is followed by the signal extension time.
Contention
window size
15 or 31 to
1,023 slots
If the station supports only 802.11b rates, it will be 31 slots for
compatibility. Otherwise, the contention window may be shorter.
Preamble
duration
20 ms
Chapter 15. A Peek Ahead at 802.11n:
MIMO-OFDM
802.11 task group N (TGn) has an interesting goal. Most IEEE task groups focus on increasing the
peak throughput, making data fly as fast as possible during the time it is being transmitted. TGn's
goal is to achieve 100 Mbps net throughput, after subtracting all the overhead for protocol
management features like preambles, interframe spacing, and acknowledgments. Although the goal
is 100 Mbps net throughput, the final proposal seems certain to blow past that number, and offer
many times that throughput in maximum configurations. There are two roads to 100 Mbps: improve
the efficiency of the MAC, increase the peak data rate well beyond 100 Mbpsor both.
Six complete proposals were made to the group creating the eventual 802.11n, but support has
coalesced around two main proposals, from groups named TGnSync and WWiSE (short for "WorldWide Spectrum Efficiency"). Both camps have chip-makers. Atheros, Agere, Marvell, and Intel are
part of TGnSync; Airgo, Broadcom, Conexant, and Texas Instruments are the core of WWiSE.
However, quite a few manufacturers of electronic devices that might use 802.11 (Cisco, Nokia,
Nortel, Philips, Samsung, Sanyo, Sony, and Toshiba) have also become part of the effort, and they
are disproportionately represented in TGnSync.
At a very high level, both proposals are similar, though they differ in the emphasis on increasing
peak data rates versus improving efficiency. Each of them makes use of multiple-input/multipleoutput (MIMO) technology in several configurations and provides for backwards compatibility with
installed systems in the same frequency band. Both support operation in the current 20 MHz
channels, with provisions to use double-width 40 MHz channels for extra throughput.
As the standards war is fought across the globe at IEEE meetings, a "pre-N" access point has already
hit the streets, based on Airgo's chipset. Purchasing it well before the standards process is underway
is a roll of the dice. When most "pre-G" products were brought to market, the task group had begun
to work in earnest on a single proposal. TGn is currently in the "dueling proposal" stage right now,
and there is no guarantee that an early device will be firmware upgradeable to the final 802.11n
standard. 802.11n is likely to be the last chance to standardize a PHY this decade. Developing a
standard is as much political engineering as technical engineering. IEEE rules require that a proposal
get a 75% supermajority vote before becoming the basis for a standard. As this book went to press,
TGnSync was garnering a clear majority of support, but was still falling short of the necessary 75%. I
expect that features from competing proposals will be incorporated into the working document to
bring the vote count to the necessary level. As a result, this chapter describes both of the main
competing proposals. Although TGnSync will probably be the basis for the 802.11n specification,
some horse trading will likely result in a few WWiSE features being incorporated.
This chapter describes both the WWiSE and TGnSync proposals. The final standard will have some
resemblance to both of them, and will likely pick and choose features from each. Fortunately, many
basic concepts are shared between the two. As you read this, keep in mind that the proposals
themselves may have changed quite a bit since the drafts upon which this chapter was based were
written.
Common Features
Although the two proposals are different, there is a great deal of similarity between the two.
Practically speaking, some features are required to reach the goal of 100 Mbps throughput.
Multiple-Input/Multiple-Output (MIMO)
Up until 2004, 802.11 interfaces had a single antenna. To be sure, some interfaces had two antennas
in a diversity configuration, but the basis of diversity is that the "best" antenna is selected. In
diversity configurations, only a single antenna is used at any point. Although there may be two or
more antennas, there is only one set of components to process the signal, or RF chain. The receiver
has a single input chain, and the transmitter has a single output chain.
The next step beyond diversity is to attach an RF chain to each antenna in the system. This is the
basis of Multiple-Input/Multiple-Output (MIMO) operation.[*] Each RF chain is capable of
simultaneous reception or transmission, which can dramatically improve throughput. Furthermore,
simultaneous receiver processing has benefits in resolving multipath interference, and may improve
the quality of the received signal far beyond simple diversity. Each RF chain and its corresponding
antenna are responsible for transmitting a spatial stream. A single frame can be broken up and
multiplexed across multiple spatial streams, which are reassembled at the receiver. Both the WWiSE
and TGnSync proposals employ MIMO technology to boost the data rate, though their applications
differ.
[*]
MIMO is pronounced "MyMoe." I attended a symposium in which a standards committee attendee described the standardization
vote on the acronym's pronunciation.
MIMO antenna configurations are often described with the shorthand "YxZ," where Y and Z are
integers, used to refer to the number of transmitter antennas and the number of receiver antennas.
For example, both WWiSE and TGnSync require 2x2 operation, which has two transmit chains, two
receive chains, and two spatial streams multiplexed across the radio link. Both proposals also have
additional required and optional modes. I expect that the common hardware configurations will have
two RF chains on the client side to save cost and battery power, while at least three RF chains will be
used on most access points. This configuration would use 2x3 MIMO for its uplink, and 3x2 MIMO on
the downlink.
Channel Width
802.11a currently uses 20 MHz channels because that is the channel bandwidth allowed by all
regulators worldwide. Doubling the channel bandwidth to 40 MHz doubles the theoretical information
capacity of the channel. Although promising for the future, some regulators do not currently allow 40
MHz operation. Japan is the most notable exception.
MAC Efficiency Enhancements
As this book has repeatedly pointed out, the efficiency of the 802.11 MAC is often poor. In most
usage scenarios, it is very difficult to exceed 50-60% of the nominal bit rate of the underlying
physical layer. Every frame to be transmitted requires a physical-layer frame header, as well as the
pure overhead of preamble transmission. The 802.11 MAC adds further overhead by requiring that
each frame be acknowledged. Overhead is particular bad for small frames, when the overhead takes
more time than the frame data itself. Figure 15-1 shows the efficiency, defined as the percentage of
the nominal bit rate devoted to MAC payload data, for a variety of frame sizes. The values in the
figure are exclusively for MAC payload data. Any network measurement would require additional LLC
data, and networks that are encrypted would have additional overhead bytes. Furthermore, most
network protocols provide their own acknowledgment facilities, which further reduces real-world
efficiency. The point of Figure 15-1 is that small frames have particularly poor efficiency.
Both TGnSync and WWiSE adopt techniques to improve the efficiency of the radio channel. Concepts
are similar, but the details differ. Both offer some form of block ACKs (sometimes called frame
bursting). By removing the need for one acknowledgment frame for every data frame, the amount of
overhead required for the ACK frames, as well as preamble and framing, is reduced. Block
acknowledgments are helpful, but only if all the frames in a burst can be delivered without a
problem. Missing one frame in the block or losing the acknowledgment itself carries a steep penalty
in protocol operations because the entire block must be retransmitted.
Figure 15-1. MAC Efficiency
Frame aggregation is also part of both proposals. Many of the packets carried by 802.11 are small.
Interactive network sessions, such as telnet and SSH, make heavy use of rapid-fire small packets.
Small packets become small frames, each of which requires physical-layer framing and overhead.
Combining several small packets into a single relatively large frame improves the data-to-overhead
ratio. Frame aggregation is often used with MAC header compression, since the MAC header on
multiple frames to the same destination is quite similar.
WWiSE
The WWiSE consortium includes several well-known chipmakers: Airgo (the manufacturer of the first
"pre-N" devices on the market), Broadcom, Conexant, and Texas Instruments. Motorola joined the
consortium in February 2005, just as this book headed to press.
MAC Enhancements
As would be expected from a name touting spectral efficiency, WWiSE is more the more heavily
weighted towards improving the MAC efficiency of the two proposals. To get to 100 Mbps net payload
throughput, 12,000 bytes (960,000 bits) need to be transmitted in 960 microseconds. WWiSE's PHY
specification has a 135 Mbps data rate in a basic two-antenna configuration with two data streams,
which can move the data in 711 microseconds. The remaining 249 microseconds are used for
preambles, framing, interframe spacing, and the single block acknowledgment.
Channels and radio modes
WWiSE uses both 20 MHz and 40 MHz channels. 40 MHz operation may be through a single 40 MHz
channel, or through a 20 MHz channel pair in which both channels are used simultaneously for data
transmission. One channel is designated as the primary channel, and operates normally. The
secondary channel is used only for channel aggregation, and does not have stations associated on it.
The secondary channel is used for "overflow" from the primary; carrier sensing functions are
performed only on the primary channel.
Although the use of two channels is really a physical layer operation, there are some housekeeping
functions performed by the MAC. A new information element, the Channel Set element, is sent in the
primary channel Beacon frames so that stations are informed of the secondary channel in the pair.
Access points also send Beacon frames on the secondary channel; unlike most Beacon operations,
though, the purpose is to discourage clients from associating, or other devices from choosing that
channel for operation. A secondary channel Beacon frame is very similar to the primary channel
Beacon, but the only supported rate is a mandatory MIMO PHY rate. To further discourage use of the
channel, it may also include the contention-free information element.
Protection
Like 802.11g, the new PHYs require enhanced protection mechanisms to avoid interfering with
existing stations. Naturally, the protection mechanisms specified in 802.11g are adopted for
operation of 2.4 GHz stations that may have to avoid interfering with older direct sequence or
802.11b equipment. When access points detect the presence of older equipment, it will trigger the
use of RTS-CTS or CTS-to-self protection as described in Chapter 14.
However, additional protection may be required to avoid having a MIMO station transmit at a rate not
understood by 802.11a or 802.11g equipment. The WWiSE proposal contains an OFDM protection
scheme to allow MIMO stations to appropriately set the NAV on older OFDM stations. The protection
mechanism is identical to the one described in Chapter 14, but it takes place using OFDM data rates.
Finally, the WWiSE proposal uses two bits in the ERP information element in Beacon frames to
indicate whether OFDM protection is needed. In some cases, OFDM protection may be needed to
assist an older 802.11g network, but no protection is needed for 802.11b stations. Access points
monitor the radio link to determine if OFDM protection is needed. To assist stations using channel
pairs, they also report on whether a secondary channel is in use.
Aggregation, bursting, and acknowledgment
The WWiSE proposal increases the maximum payload size from 2,304 bytes to over 8,000 bytes.
Increasing the payload increases the payload-to-overhead and the ratio can increase efficiency if the
larger frames or bursts can be delivered successfully.
Aggregation bundles multiple higher-level network protocol packets into a single frame. Each packet
gets a subframe header with source and destination addresses, and a length to delimit the packet, as
shown in Figure 15-2. Aggregation can only be used when the frames bundled together have the
same value for the Address 1 field, which is the receiver of the frame. Frames from an access point in
an infrastructure network use Address 1 as the destination, so access points can only aggregate
frames bound for a single station. A station in an infrastructure network can, however, aggregate
frames to multiple destinations. Station transmissions use the Address 1 field for the AP, since all
frames must be processed by the AP prior to reaching the backbone network. Upon aggregation, the
destination address is the "next hop" processing station, and the source is the creator of the frame.
Upon deaggregation, the individual subframes will be processed according to the sub-frame headers.
Due to the requirement that the receiver address must be the same, it is not possible to aggregate a
mixture of unicast, broadcast, and multicast data. The proposal contains no rules about when to use
aggregation.
Figure 15-2. Aggregation in WWiSE
Bursting is a related, but slightly different concept. Frame aggregation glues higher-layer protocol
packets together for transmission in larger lumps. Bursting does the same at the physical layer. Once
a station has invested a significant amount of protocol overhead to obtain control of the channel, it
can just keep on transmitting. One of the advantages of using multiple physical frames, as opposed
to higher-layer frames, is that each physical frame has its own source and destination. A frame burst
can consist of traffic intended for a variety of different destination addresses. In a frame burst, there
are two additional interframe spaces defined, the Zero Interframe Space (ZIFS) and the Reduced
Interframe Space (RIFS). Successive frames that use the same transmit power may use the ZIFS for
immediate transmission. If the transmit power is changed between frames, the RIFS may be used.
The RIFS is shorter than other interframe spaces, though, so it allows a station to retain control of
the channel. In Figure 15-3, the first frame cannot be aggregated, and is transmitted after the
transmitter gains control of the channel. Once it has gained control, it can hold on as long as allowed.
The second and third frames use the same transmission power, and so are transmitted after the zero
interframe space. Additionally, they share the Address 1 field and are therefore bundled into an
aggregate frame. For transmission of the next frame, power needs to be changed, requiring the use
of the reduced interframe space. The fourth and fifth frames can be aggregated, and are transmitted
as a single aggregate frame. When the queued data has been transmitted, the station relinquishes
control of the channel.
Figure 15-3. Bursting in WWiSE
In the initial version of the 802.11 MAC, a positive acknowledgment was required for every unicast
data frame. WWiSE lifts this restriction, and allows for a more flexible acknowledgment policy. In
addition to the "normal" policy, frames can be transmitted without an acknowledgment requirement,
or with block acknowledgments instead.
The WWiSE MIMO PHY
The WWiSE proposal is a slight evolution of 802.11a, using MIMO technology. The basic channel
access mechanisms are retained, as is the OFDM encoding. At a high level, the WWiSE PHY is mainly
devoted to assigning bits to different antennas.
Structure of an operating channel
Like 802.11a, the radio channel is divided into 0.3125 MHz subcarriers. As in the 802.11a channel
subdivisions, a 20 MHz channel in the WWiSE proposal is divided into 56 subcarriers. 40 MHz
channels, which are optional, are divided into 112 subcarriers. In addition to being optional, 40 MHz
channels are only supported in the 5 GHz band because it is not possible to squeeze multiple 40 MHz
channels into the ISM band. (And if you thought network layout was hard with three channels, wait
until you try with two!) Figure 15-4 shows the structure of both the 20 MHz and 40 MHz operating
channels in the WWiSE proposal.
As in 802.11a, subcarriers are set aside as pilots to monitor the performance of the radio link. Fewer
pilot carriers are needed in a MIMO system because the pilot carriers run through as many receiver
chains. A 20 MHz 802.11a channel uses four pilot
Figure 15-4. WWiSE pilot carrier structure
subcarriers. In the WWiSE proposal, a 20 MHz channel requires only two pilot carriers because each
pilot is processed by two receiver chains, which has the same effect as four pilots processed by a
single receiver chain. With fewer pilots, more subcarriers can be devoted to carrying data. 20 MHz
WWiSE channels have 54 data subcarriers; 40 MHz channels have exactly twice as many at 108.
Modulation and encoding
The WWiSE proposal does not require new modulation rates. It uses 16-QAM (4 bit) and 64-QAM (6
bit) modulation extensively, but does not require finer-grained modulation constellations.
Coding is enhanced, however. A new convolutional code rate of 5/6 is added. Like the 2/3 and 3/4
code rates defined by 802.11a, the 5/6 code is defined by puncturing the output to obtain a higher
code rate. WWiSE also defines the use of a low density parity check (LDPC) code.
Interleaver
In 802.11a, the interleaver is responsible for assigning bits to subcarriers. MIMO interleavers are
more complex because they must assign bits to a spatial stream in addition to assigning bits to
positions on the channel itself. The WWiSE interleaver takes bits from the forward error coder and
cycles through each spatial stream. The first bit is assigned to the first spatial stream, the second bit
is assigned to the second spatial stream, and so on. The interleaver is also responsible for scrambling
the encoded bits within each spatial stream.
Space-time block coding
In most cases, an antenna will be used for each spatial stream. However, there may be cases when
the number of antennas is greater than the number of spatial streams. If, for example, most APs
wind up using three antennas while clients only use two, there is an "extra" transmit antenna, and
the two spatial data streams need to be assigned to the three antennas. Transmitting a single spatial
stream across multiple antennas is called space-time block coding (STBC).
The basic rule for splitting a spatial stream across multiple antennas is to transmit two related
streams on different antennas. As discussed in Chapter 13 on 802.11a, the radio wave is composed
of in-phase and quadrature components, where the quadrature wave is a quarter-cycle out of phase
with the in-phase component. Phase shifts are represented mathematically by the imaginary part of
the complex number in the constellation. The complex conjugate of a complex number has the same
real part, but flips the sign on the imaginary part. Physically, the radio wave from the complex
conjugate will have the same in-phase component, but the quadrature component will have the
oppose phase shift. When there are extra antennas, the WWiSE proposal mandates that a spatial
stream and its complex conjugate are transmitted on an antenna pair. Table 15-1 reviews the rules.
The rules for splitting spatial streams are independent of the channel bandwidth, although 40 MHz
spatial streams will carry more bits.
Table 15-1. WWiSE encoding rules when antennas outnumber spatial
streams
Transmit
antennas
Spatial
streams
First spatial stream Second spatial stream
Third spatial
stream
2
1
Coded across
antennas 1 and 2
N/A
N/A
3
2
Coded across
antennas 1 and 2
Transmitted normally on
third antenna
N/A
4
2
Coded across
antennas 1 and 2
Coded across antennas 3
and 4
N/A
4
3
Coded across
antennas 1 and 2
Third antenna
Fourth antenna
Modulation rates
There are 24 data rates defined by the WWiSE PHY, with 49 different modulation options. Rather
than take up a great deal of space in a table, here is a basic formula for the data rates:
Data rate (Mbps) = 0.0675 x channel bandwidth x number of spatial streams x coded bits per
subcarrier x code rate
Channel bandwidth
Either 20 for 20 MHz channels, or 40 for 40 MHz channels or channel pairs.
Number of spatial streams
The number of spatial streams can be equal to 1, 2, 3, or 4. It must be less than or equal to
the number of transmission antennas. Support for at least two spatial streams is mandatory.
Coded bits per subcarrier
In most cases, this will either be 6 for 64-QAM or 4 for 16-QAM. BPSK (1 coded bit per
subcarrier) and QPSK (2 coded bits per subcarrier) are only supported in the 20 MHz channel
mode with one spatial stream.
Code rate
The code rate may be 1/2 or 3/4 when used with 16-QAM, and 2/3, 3/4, or 5/6 when used with
64-QAM.
There may be multiple ways to get to the same data rate. As an example, there are four ways to get
108 Mbps:
Four spatial streams in 20 MHz channels, using 16-QAM with R=1/2.
Two spatial streams in 20 MHz channels, using 64-QAM with R=2/3.
One spatial stream in a 40 MHz channel, using 64-QAM with R=2/3.
Two spatial streams in 40 MHz channels, using 16-QAM with R=1/2.
In a basic mode with a single spatial stream, channel capacity is slightly higher than with 802.11a
because fewer pilot carriers are used. Single-channel modulation tops out at 60.75 Mbps, rather than
the 54 Mbps in 802.11a. By using all the highest throughput parameters (four 40 MHz spatial
streams, with 64-QAM and a 5/6 code), the WWiSE proposal has a maximum throughput of 540
Mbps.
MIMO and transmission modes
Previous 802.11 PHY specifications had fairly simple transmission modes. The WWiSE proposal has
14 transmission modes, depending on 3 items:
The number of transmit antennas, noted by xTX, where x is the number of transmit antennas. It
ranges from 1 to 4, although a single antenna is only supported for 40 MHz channels. All 20
MHz channels must use at least two transmit antennas, though they may have only one spatial
stream.
Whether the frame is used in a greenfield (GF) or mixed mode (MM) environment. Mixed mode
transmissions use physical headers that are backwards-compatible with other OFDM PHYs, while
greenfield transmissions use a faster physical header.
The channel bandwidth, which may be 20 MHz or 40 MHz.
Table 15-2 shows the resulting 14 transmission modes. There are several physical layer encodings
defined for each of these modes, and they will be discussed in the PLCP section. The number of
active antennas is only loosely related to the number of spatial streams. A system operating in the
4TX40MM mode has four transmit antennas, but it may have two or three spatial streams.
Table 15-2. WWiSE transmission modes
Greenfield
20 MHx channels
40 MHz channels
2TX20GF
1TX40GF
3TX20GF
2TX40GF
4TX20GF
3TX40GF
4TX40GF
Mixed mode
2TX20MM
1TX40MM
3TX20MM
2TX40MM
4TX20MM
3TX40MM
4TX40MM
WWiSE PLCP
The PLCP must operate in two modes. In Greenfield mode, it operates without using backwardscompatible physical headers. Greenfield access is simpler: it can operate without backwards
compatibility. As a starting point, consider Figure 15-5; it shows the PLCP encapsulation in the
1TX40GF, 2TX20GF, and 2TX40GF modes.
Figure 15-5. Greenfield 1TX40 and 2TX20/2TX40 modes
The fields in the frame are similar in name and purpose to all of the other PLCP frames discussed in
this book.
MIMO-OFDM PLCP Preamble
The preamble consists of well-known bit sequences to help receivers lock on to the signal.
Depending on the transmission mode, the preamble may be split into multiple parts. It
generally consists of both short and long training sequences. In the WWiSE proposal, the same
preamble is transmitted on all the antennas, but with small time shifts relative to the others.
Figure 15-5 shows the training sequences used by two-antenna transmission modes. Although
the training sequences consist of different bits, the shift is the same. Naturally, the single
antenna 40 MHz mode would only have one active antenna transmitting a preamble.
SIGNAL-N
The SIGNAL-N field contains information that helps to decode the data stream. It is always
sent using QPSK, R=1/2, and is not scrambled. It contains information on the number of
spatial streams, channel bandwidth, modulation, and coding, and a CRC. More detail on the
SIGNAL-N field follows this section.
SERVICE
The SERVICE field is identical to its usage in 802.11a. Unlike the other components of the PLCP
header, it is transmitted in the Data field of the physical protocol unit at the data rate of the
embedded MAC frame. The first eight bits are set to 0. As with the other physical layers, MAC
frames are scrambled before transmission; the first six bits are set to 0 to initialize the
scrambler. The remaining nine bits are reserved and must set to 0 until they are adopted for
future use.
Data
The final field is a sequence of four microsecond symbols that carry the data. Data bits have six
zero tail bits to ramp down the error correcting code, and as many pad bits as are required to
have an even symbol block size.
The SIGNAL-N field
The SIGNAL-N field is used in all transmission modes. It has information to recover the bit stream
from the data symbols. The SIGNAL-N field is shown in Figure 15-6.
Figure 15-6. WWiSE SIGNAL-N field
CONFIG
Six fields are grouped into the Configuration subfield.
NSS (number of spatial streams)
Three bits are used to indicate how many spatial streams are used. The value is zerobased, so it ranges from zero to three.
NTX (number of transmission antennas)
Three bits are used to indicate how many antennas are used to carry the number of
spatial streams. The value is zero-based, so it ranges from zero to three.
BW (bandwidth)
Two bits carry the channel bandwidth. 20 MHz is represented by zero, and 40 MHz is
represented by one.
CR (code rate)
Three bits indicate the code rate. 1/2 is zero, 2/3 is one, 3/4 is two, and 5/6 is three.
CT (code type)
Two bits indicate the type of code. Zero is a convolutional code, and one is the optional
LDPC.
CON (constellation type)
Three bits indicate the type of constellation: zero for BPSK, one for QPSK, two for 16QAM, and three for 64-QAM.
LENGTH
A 13-bit identifier for the number of bytes in the payload of the physical frame. It ranges from
zero to 8,191.
LPI (Last PSDU indicator)
When multiple physical frames are sent in a burst, the LPI bit is set on the last one to notify
other stations that the burst is coming to an end.
CRC
The CRC is calculated over all the fields except for the CRC and the tail bits.
Tail
Six bits are used as tail bits to ramp down the convolutional coder.
In the other transfer modes, shown in Figure 15-7, the preamble is split into chunks. In between the
chunks, there may be Signal fields. SIGNAL-N fields are defined by the 802.11n proposal and are
only decoded by 802.11n stations; the SIGNAL-MM field is used to retain backwards compatibility in
a mixed mode with older OFDM stations. It is identical to the Signal field used by 802.11a, and is
shown in Figure 13-16.
Figure 15-7. PLCP frame format for other transfer modes
WWiSE PMD
Figure 15-8 shows the basic layout of the WWiSE transmitter. It is essentially the same as the
802.11a transceiver, but it has multiple transmit chains. The interleaver is responsible for dividing
coded bits among the different transmit chains and spatial streams.
Figure 15-8. WWiSE transceiver
Sensitivity is specified by the proposal, and it is identical to what is required of 802.11a receivers.
Table 15-3 shows the required sensitivity. The proposal does not have any adjacent channel rejection
requirements.
Table 15-3. WWiSE receiver sensitivity
Constellation Rate Sensitivity (dBm)
802.11a Sensitivity (dBm), for reference
BPSK
1/2
-82
-82
BPSK
3/4
-81
-81
QPSK
1/2
-79
-79
QPSK
3/4
-77
-77
16-QAM
1/2
-74
-74
16-QAM
3/4
-70
-70
64-QAM
2/3
-66
-66
64-QAM
3/4
-65
-65
Constellation Rate Sensitivity (dBm)
802.11a Sensitivity (dBm), for reference
64-QAM
N/A
5/6
-64
Characteristics of the WWiSE PHY
Parameters specific to the WWiSE PHY are listed in Table 15-4. Like the other physical layers, it also
incorporates a number of parameters to adjust for the delay in various processing stages in the
electronics.
Table 15-4. WWiSE MIMO PHY parameters
Parameter
Value
Maximum MAC frame
length
8,191 bytes
Slot time
9 ms
SIFS time
16 ms
RIFS time
2 ms
Contention window
size
15 to 1,023
slots
Preamble duration
16 ms
PLCP header duration
4 ms
Receiver sensitivity
-64 to -82
dBm
Notes
The SIFS is used to derive the value of the other interframe
spaces (DIFS, PIFS, and EIFS).
Depends on speed of data transmission
TGnSync
The TGnSync consortium is composed of a wider array of companies. In addition to the chipmakers
that one would expect to find (Atheros, Agere, and Intel, and Qualcomm), TGnSync notably includes
manufacturers of other electronic devices. Network equipment manufacturers and even consumer
electronics companies are represented. One of the goals of TGnSync is to support new networked
devices in the home; promotional materials refer to sending HDTV or DVD video streams across
wireless networks. The goal of streaming video probably accounts for some of the emphasis placed
on high peak data rates.
TGnSync MAC Enhancements
Although the TGnSync proposal has a higher peak data rate, the group did not completely neglect the
development of MAC enhancements to improve efficiency and operation. Efficiency is improved
through the development of frame aggregation and bursting, as well as changes to acknowledgment
policies. Some protection of older transmissions is performed at the MAC layer. Notably, several MAC
enhancements are designed to save battery power, which is likely a reflection of the group's
membership.
Channels, radio modes, and coexistence
Although some regulators do not allow them, the TGnSync proposal makes 40 MHz channel support
mandatory. If it were adopted without change, a TGnSync chipset would support both 20 MHz and 40
MHz channels, even in regulatory domains that did not allow the latter channel bandwidth. The
TGnSync proposal also has MAC features that enable the use of networks with both 20 MHz-and 40
MHz-capable stations. When stations have large amounts of data to transmit, it is possible to
negotiate a temporary use of a wider channel before falling back to 20 MHz operation.
MAC operational modes can also be classified based on the types of stations in the network. Pure
mode networks consist only of 802.11n stations. No protection is necessary to account for older
802.11a and 802.11g stations. Alternatively, TGnSync 802.11n devices may operate in the legacy
mode just like an 802.11a or 802.11g station. Most operation, though, will be in mixed mode, where
a TGnSync network must co-exist with a legacy network on the same channel, and may accept
associations from older 802.11a or 802.11g stations.
Association requests are handled differently in each mode. Pure mode networks stay pure by ignoring
association requests from older stations, and sending Beacon frames with an information element
that directs associated stations to use only the new 802.11n transmission modes. Pure mode
networks also transmit Beacon frames using the TGnSync high throughput PLCP, which makes them
unreadable by legacy devices. Mixed mode access points are visible to legacy devices because they
transmit Beacons using the legacy format.
Mixed mode is required to coexist with older devices. (If the experience of 802.11g deployment is
any guide, most 802.11n devices are likely to operate in the mixed mode for quite some time.) Mixed
mode is a broad classification with several subdivisions. Mixed capable networks will allow association
from legacy devices, but do not divide time between legacy and high-throughput transmissions.
Access points in managed mixed networks do actively divide the time between high-throughput
transmissions and legacy transmissions. Much like the division between the contention-free period
and the contention period (see Chapter 9), an AP operating in managed mixed mode will allow legacy
stations their timeslice, while using mechanisms similar to the protection mechanism to reserve some
timeslice for MIMO stations only.
Aggregation and bursting
Initial 802.11 stations typically send frames in the order they are received. For throughput purposes,
it is highly desirable to reorder frames so that they can coalesce into larger aggregated frames.
Aggregation in TGnSync is a MAC-layer function that bundles several MAC frames into a single PLCP
frame for transmission.
Figure 15-9 shows the basic format of a single physical-layer frame containing several MAC layer
frames. Several MAC frames are put into the same PLCP frame, with an appropriate delimiter
between them. The delimiter has a small reserved field, a length field for the following MAC frame, a
CRC to protect the delimiter, and a unique pattern to assist in recovering individual frames from the
aggregate. MAC frames are put into the aggregate without modification, and contain the full header
and MAC CRC. Even if one frame out of an aggregate is lost, it may be possible to successfully
receive all the remaining frames.
Figure 15-9. TGnSync frame aggregation
Exchanging aggregated frames is only possible once the channel has been configured for it. Figure
15-10 illustrates the process. The sender of an aggregate, called the initiator, must send an Initiator
Aggregation Control (IAC) frame. IAC frames work much like RTS frames, but have additional fields
to assist with channel control. Initiators can request channel measurements, offer different types of
coding on the aggregate frame, and accept aggregates in the return direction. Upon receiving the
IAC, the destination system, called the responder, generates a Responder Aggregation Control (RAC)
frame. RAC frames work much like CTS frames: they close the loop by notifying the sender that an
aggregate will be accepted, and finishing the parameter negotiation. When aggregate frames are
received, an acknowledgment is required. TGnSync defines a new acknowledgment type, the
BlockACK, which can be used to acknowledge all the MAC frames contained in an aggregate.
To further improve MAC efficiency, TGnSync defines a MAC header compression algorithm for use in
conjunction with aggregate frames. It works in the same manner as Van Jacobsen header
compression on serial dial-up lines. Frames between two destinations share most of the fields in the
MAC header, most notably the MAC addresses inside the packet. Therefore, a one-byte Header ID
(HID) is assigned to a unique set of the three MAC addresses inside a MAC frame. The Header ID can
also save the Duration field, since the aggregate will have its own Duration, as well as the two bytes
for QoS control. When frames are transmitted between the same sender
Figure 15-10. TGnSync block acknowledgment
and destination, rather than repeating the same 22 bytes of header information, it is replaced by the
corresponding single byte Header ID. Figure 15-11 (a) shows the use of the header compression MAC
frames. First, a header frame containing the full header is transmitted, and a header ID is assigned.
The header ID can be used to reference the prior full header by sending a single byte to reference
previously-transmitted information about the Duration, addressing, and QoS data.
Figure 15-11 (b) illustrates the use of header compression. Five frames for two destinations from the
MAC have been aggregated into a single frame for physical transmission. Rather than transmit a
complete header on each constituent MAC frame in the aggregate, the system uses header
compression. There are two destinations and therefore two unique MAC headers. They are each
transmitted and assigned a header ID number. The five data frames following the aggregate each
refer to the appropriate header number. A header ID number is unique only within the context of a
single aggregate frame. Compared to transmitting full headers on all five frames, the overhead due
to MAC framing is cut by more than half.
Header compression is useful when a single aggregate contains multiple frames between the same
source and destination pair. However, the benefits of aggregation in TGnSync are not confined to
pairs. Single-receiver aggregation is required; an optional extension allows aggregate frames to
contain MAC frames for multiple receivers, in which case they are called Multiple Receiver Aggregate
(MRA) frames. Inside the single transmitted aggregate frame, there are multiple Initiator Access
Control frames. Each IAC specifies an offset to transmit the response to the aggregated frames,
which will usually be a block acknowledgment response. To distinguish multiple receiver aggregate
frames from single-receiver aggregate frames, multiple-receiver frames start with a control item
called the Multiple Receiver Aggregate Descriptor (MRAD). Figure 15-12 shows the operation of
multiple-receiver aggregation. The initiator's aggregate frame starts with the aggregate descriptor,
and is followed by the aggregated frames for each destination. An IAC frame is used to divide them.
Figure 15-11. TGnSync MAC header compression
Figure 15-12. TGnSync MRA
Protection
As with all PHYs that have followed existing hardware on to the market, the TGnSync proposal
implements protection to avoid having the new PHY step on transmissions from the old PHY.
Protection in the TGnSync proposal can take one of two main forms. The first class is based on the
MAC's virtual carrier sensing mechanism with the network allocation vector. The second class is
based on "spoofing," which uses the existing PLCP header format to carry duration information. Each
station may make its own determination as to the appropriate mechanism.
Setting long NAV values to protect the duration of a frame exchange is a small adaptation of the
802.11g protection mechanism. At the start of a frame exchange, the RTS frame will contain a NAV
long enough to protect the entire frame exchange. The RTS frame is sent using a "legacy" rate, and
can be understood by existing OFDM receivers. In response, the target station sends a CTS message
back, also with a long NAV value. According to the basic access rules of the MAC, other stations defer
access to the medium due to the RTS/CTS clearing, and the two stations are free to exchange frames
at higher data rates using modulations that would not be understood by older stations. LongNAV
intervals may be terminated early by using a CF-End frame. When this protection is used for
aggregate frames, the RTS is replaced by an Initiator Access Control (IAC) frame, and the CTS is
replaced by a Responder Access Control (RAC) frame; the principle of operation, however, remains
identical. See Figure 15-13.
Figure 15-13. TGnSync protection: LongNAV
The second class of protection is called "spoofing," and depends on setting the length field in the
PLCP header. TGnSync retains the existing OFDM header described in Chapter 13. Because it is
identical to the 802.11a/g format, spoofing is effective with all stations. The OFDM PLCP header
contains two numbers that are used by receivers to determine how long the transmission will take.
The SIGNAL field, which is shown in Figure 13-16, encodes both the transmission rate for the body
and its length in bits. Stations decode the signal field and divide the number of bits by the rate to
come up with an approximate transmission time.[*] To maximize the amount of time that can be
spoofed, the data rate in the legacy SIGNAL field is always set to the lowest possible value of six
Mbps.
[*]
There are some slight offsets to account for interframe spacing, but the concept remains identical.
In pairwise spoofing, two stations will each send an incorrect length and rate so that older stations
will be in receiving mode for the duration of the current frame and its next response. Newer TGnSync
stations ignore the older SIGNAL field, and use an 802.11n SIGNAL field instead. Figure 15-14
illustrates pairwise spoofing. When Frame 1 is transmitted, pairwise spoofing is used to lengthen the
receiving time of the frame through the end of Frame 2. TGnSync stations will interpret the spoofing
as a longer NAV, and will therefore act as if the NAV were set for the duration of Frame 2. Naturally,
a station within range of the responder will be set to the receiving state; if, however, there is a
hidden node, the NAV will protect transmission over Frame 2. 802.11a/g stations will interpret the
spoofed time as receiving time, even if they are out of range of the second frame. When Frame 2 is
transmitted, it also employs pairwise spoofing to protect Frame 2 and Frame 3.
Figure 15-14. TGnSync protection: pairwise spoofing
If a long lock-out period is needed for multiple responses to a single frame, single-ended spoofing
may also be used. With single-ended spoofing, the first frame in the exchange uses spoofing to
protect the entire exchange, allowing all the responses to come in during the protected period. Figure
15-15 illustrates single-ended spoofing with frame aggregation. The first aggregate frame is a
multiple-receiver aggregate, allowing responses from two other stations. It sets spoofed duration
equal to the time expected for the entire exchange. TGnSync stations will go into receiving mode for
the duration of the first frame, and then act as if the NAV were set for the spoofed duration. Legacy
devices go into receiving mode for the entire spoofed duration.
Figure 15-15. TGnSync protection: single-ended spoofing
Powersaving
TGnSync defines the Timed Receive Mode Switching (TRMS) protocol to conserve energy and extend
battery life. Traditional 802.11 powersaving works by completely shutting down an interface and
requiring buffering at the AP. In single-input/single-output radios, there is only one RF chain to shut
down. With MIMO systems, however, there can be significant power savings by shutting down
unused RF chains, but retaining a single active chain to monitor the radio link. The two states of the
system are called MIMO enabled for full receive capability, and MIMO disabled when all but one RF
chain is shut down.
Stations activate TRMS power saving by including an information element in the association request.
The basic parameter in TRMS powersaving is the hold time. After a station transmits a frame, it stays
awake for the duration of the hold time. Any transmitted frame resets the hold timer to its maximum
value. Setting the hold time to zero indicates that the station will remain fully operational for one slot
time before sleeping.
In infrastructure networks, the AP is responsible for maintaining the TRMS hold timer for every
station. If the timer elapses, the AP must conclude it has entered the MIMO disabled state, and
trigger it to power on sleeping receive chains. In an independent BSS, each station must maintain a
hold timer for all the other stations.
The timer is a tunable parameter. If it is set to a larger value, stations will use more power to keep
the receiver fully operational. Throughput is likely to be better, but at the cost of some battery life. In
some cases, network capacity may not be affected much at all. Networks that use the NAV
lengthening procedure for protection must transmit the initial RTS/CTS exchange in single-antenna
mode manner compatible with all OFDM stations, and will cause stations to enable MIMO operation
without additional frames. Advanced transmission modes may suffer, however, because many of
them require multiantenna frame exchanges to become fully operational.
TGnSync PHY Enhancements
To develop a higher peak data rate, the TGnSync proposal depends on technology similar to WWiSE.
Frames are divided into spatial streams that can be multiplexed across antennas in a MIMO
configuration. More aggressive coding, including a larger constellation, higher convolutional code
rate, and a reduced guard interval are present to improve the data rate. Wider channels are also
required by TGnSync where supported. Support for 40 MHz channels must be built in to TGnSynccompliant devices, whereas it is optional in WWiSE.
Structure of a channel
Both 20 MHz and 40 MHz channels are divided into 0.3125 MHz subcarriers, just as in 802.11a. The
20 MHz channel is identical to an 802.11a channel, and is shown in Figure 15-16 (a). The 40 MHz
channel proposed in TGn Figure 15-16 (b) is a modification of the 20 MHz structure. Two 20 MHz
channels are bonded together, and the resulting spectral band is divided into 128 subchannels. The
center frequencies of the old 20 MHz channels are located at +/-32. The legacy channels apply a
spectral mask from -6 to +6 and roll off the amplitude of transmissions at the end of the bands. With
a single continuous channel, however, there is no need to use a spectral mask, and the middle of the
band can be used at full strength. Full-strength transmissions in the middle of the band allow for
eight new subcarriers. Using a single contiguous 40 MHz block of spectrum reclaims subcarriers that
would have otherwise been wasted. Thus, in TGnSync, a 40 MHz channel provides throughput equal
to 2.25 times the 20 MHz channel, rather than simply doubling the throughput. To further boost
throughput, one of the pilot carriers from the 20 MHz channel is removed, so a 40 MHz channel has 6
pilot carriers instead of 8.
Figure 15-16. TGnSync channel structure
Basic MIMO rates
There are 32 modulation and coding pairs defined by the TGnSync PHY. In the basic MIMO mode,
every spatial stream must use an identical modulation technique, so the data rate is simply a
multiple of the single spatial stream data rate. Rather than take up a great deal of space in a table,
here is a basic formula for the data rates:
Data rate (Mbps) = 12 x channel bandwidth factor x number of spatial streams x coded bits per
subcarrier x code rate x guard interval factor
Channel bandwidth factor
20 MHz channels are the baseline, and are assigned a channel bandwidth factor of 1. 40 MHz
channels carry more than twice the data, and are assigned a channel bandwidth factor of 2.25.
Number of spatial streams
The number of spatial streams can be equal to 1, 2, 3, or 4. It must be less than or equal to
the number of transmission antennas. Support for at least two spatial streams is mandatory.
Coded bits per subcarrier
This is either 6 for 64-QAM, 4 for 16-QAM, 2 for QPSK, or 1 for BPSK.
Code rate
The code rate may be 1/2 when used with BPSK; 1/2 or 3/4 when used with QPSK or 16-QAM;
or 2/3, 3/4, or 7/8 when used with 64-QAM.
Guard interval factor
The basic guard interval is 800 ns, and is assigned the factor 1. 400 ns guard intervals increase
throughput slightly, and are assigned a factor of 1.11.
In a basic mode with a single spatial stream, channel capacity is identical to 802.11a, with the
exception that a 7/8 rate code may be used for a 63 Mbps data rate. By using the highest capacity
parameters (four 40 MHz channels, 64-QAM with a 7/8 code rate, and the short guard interval), the
TGnSync proposal has a maximum throughput of 630 Mbps.
Transmit modes
There are three MIMO modes that the TGnSync proposal calls for. In the mandatory basic MIMO
mode, the number of spatial streams is equal to the number of antennas. Each spatial stream is
modulated and transmitted identically. Each channel is coded using the same modulation, and sent
with the same transmission power. Any changes in transmission rate are based on the implicit
feedback of lost acknowledgments.
Two optional modes take advantage of information learned about the radio channel, which is referred
to as "closed-loop" operation. TGnSync devices send "sounding" frames to each other to measure the
performance of the link. Based on the information gleaned from sounding and calibration,
beamforming can be used to boost signal quality. Higher signal quality means that a given data rate
can be used at longer range. For a given signal-to-noise ratio, a beamformed transmission can carry
more data. Beamforming is an optional protocol feature. It is unlikely that most client devices will be
unable to send beamformed transmissions. However, client devices must be able to receive
beamformed frames to receive the benefits.
In the basic MIMO with beamforming mode, every channel must be coded the same way. Before
beginning transmission, a sounding exchange is required to calibrate the radio channel. Based on the
information from the sounding exchange, the power and coding for the spatial streams is selected.
Basic beamforming mode requires that all spatial streams be transmitted at the same power with the
same coding. Basic beamforming can be used whenever the number of spatial streams is less than or
equal to the number of transmission antennas, but its signal processing advantages are most evident
when the number of antennas transmitting a signal is greater than the number of receiving antennas.
If the number of spatial streams is less than the number of transmit antennas, a spatial steering
matrix is used to assign bits to transmission antennas.
An optional advanced beamforming MIMO (ABF-MIMO) mode is also defined. It works in a manner
similar to the basic beamforming mode, but with the additional capability of using different
transmission power on each transmit stream, as well as the possibility of using a different modulation
and code on each spatial stream. Like the basic beamforming mode, it requires the gathering of radio
status information to calibrate the channel. An optional mode in the advanced beamforming mode
allows beamforming to occur in both directions if it is supported in both directions. The advanced
beamforming MIMO mode also includes one new constellation: 256-QAM, which transmits 8 coded
bits per subcarrier.
To obtain the throughput for the advanced beamforming mode, use the equation in the previous
section for each spatial stream, and add the spatial streams together. For 256-QAM, use 8 coded bits
per subcarrier. 256-QAM is only used with a rate R=3/4 code rate.
Optional coding
In addition to the convolutional code supported by the original OFDM specification, the TGnSync
proposal also includes two optional additional error correction codes. The first technique uses the
Reed-Solomon block code, which was developed in 1960. It is widely used in many digital
applications, most notably as the error-correction code on CDs and DVDs. The TGnSync proposal
combines the Reed-Solomon code with the existing convolutional code in a conventional manner.
First, the data stream is encoded with the Reed-Solomon code, and then the output of that encoding
process is handed to a convolutional code.[*] Both codes have complementary properties.
Convolutional codes work by spreading errors out over time, and can deal with relatively isolated
errors; the Reed-Solomon code works well at correcting error bursts. An alternative to the ReedSolomon/convolutional combination is a low-density parity check (LDPC) code.
[*]
The combination of a block encoder followed by a convolutional encoder is especially popular, and has been used extensively
by deep-space probes. Galileo, Cassini, the Mars Pathfinder, and the Mars Rover all used Reed-Solomon/convolutional code
combinations.
Optional short guard interval
To further improve MAC efficiency, the TGnSync proposal allows the use of a short guard interval. In
the 802.11a and 802.11g standards as well as the WWiSE proposal, the guard interval is 800 ns. In
Chapter 13, it was discussed that that the guard interval should be two to four times the delay
spread. An 800 ns guard interval allows a 200 ns delay spread, which is much higher than was
observed in many environments. Most offices and homes have a much smaller delay spread, on the
order of 50-100 ns. In that case, using a 400 ns guard interval can boost throughput by
approximately 10%.
TGnSync Physical Transmission (PLCP and PMD)
The basic frame format of the PLCP in the TGnSync proposal is shown in Figure 15-17. It uses the
same header as the existing OFDM, and therefore does not require the use of high-overhead
protection to avoid interfering with 802.11a or 802.11g networks. Fields prefaced with "L-" in the
figure are legacy fields common to 802.11a and 802.11g; see Chapters 13 and 14 for details.
Figure 15-17. TGnSync PLCP frame format
Legacy header
The first three fields in the TGnSync PLCP frame are identical to the 802.11a/g PLCP header.
L-STF (Legacy Short Training Field)
The legacy short training field is identical to its definition in 802.11a. It lasts eight
microseconds.
L-LTF (Legacy Long Training Field)
The legacy long training field is identical to its definition in 802.11a. It also lasts eight
microseconds.
L-SIG (Legacy Signal)
These three fields are identical to the corresponding fields used by 802.11a and adopted by
802.11g. For details, see Figure 13-16. The L-SIG field is transmitted using BPSK, R=1/2
modulation and coding.
The contents of the L-SIG field is ignored by TGnSync stations. When spoofing protection is
employed, the contents of the legacy signal field are not descriptive. TGnSync stations will look in the
high-throughput headers following the legacy headers for the actual length and coding rate of the
enclosed MAC frame.
To take advantage of the spatial diversity that is the result of having multiple antennas transmitting
the same legacy header, TGnSync offers an optional cyclic delay. Each antenna transmits its legacy
frame with a slight alteration in its cyclic prefix length, such that the total delay shift is 50 ns.
When 40 MHz channels are used, as in Figure 15-17 (b), the legacy header is transmitted on each 20
MHz subchannel. That is, subcarriers -58 to -6 in the lower 20 MHz subchannel and subcarriers +6 to
+58 in the upper 20 MHz subchannel both transmit the older 802.11a-style header.
High Throughput header
Immediately following the legacy preamble is a "high throughput" header specific to the TGnSync
proposal. The main component of the high throughput header is the high throughput signal (HT-SIG)
field, which is shown in Figure 15-18. The HT-SIG field is used to detect whether a frame carries
TGnSync-encoded data at high data rates, or if it is merely a legacy data frame. The HT-SIG field is
modulated conservatively, using Q-BPSK, R=1/2 modulation and coding. Q-BPSK uses two data
points in its constellation, but they are present on the in the quadrature component. The Q-BPSK
constellation is compared to the BPSK constellation in Figure 15-18 (b).
The three-byte high-throughput header is composed of several fields, and is transmitted in the order
shown. Least significant bits in each field are transmitted first.
Figure 15-18. TGnSync HT-SIG field
HT-Length (18 bits)
This field is the number of bytes in the payload of the PLCP frame. When aggregation is used, it
may be quite large if several full-size MAC frames are included in the payload.
Modulation and Coding Set (MCS) (6 bits)
One of the disadvantages to using MIMO is that there are myriad options when you account for
different modulation schemes, differing numbers of spatial streams, and different code rates.
The MCS field selects the modulation and coding scheme, along with the number of spatial
streams. Values of 0-31 are used for basic MIMO modes, and values of 33-63 are used for
advanced MIMO modes.
Advanced Coding (2 bits)
This two-bit field is used to indicate whether the optional advanced coding is used. No
advanced coding is indicated by zero. LDPC is indicated by 1. Reed-Solomon coding is indicated
by 2. The value 3 is not used.
Sounding packet (1 bit)
Requests and responses used to measure channel performance set this bit. When set, it
indicates that every antenna is transmitting its own spatial stream. If it is not set, then the
frame should not be used to measure channel information.
Number of HT-LTFs (2 bits)
Following the HT-Signal field are high-throughput training fields. Each spatial stream requires a
training field.
Short Guard Interval (1 bit)
When set to one, this bit flag indicates that the short 400 ns guard interval is used on MIMO
symbols in the Data field of the frame.
Aggregation (1 bit)
If this bit is set to one, it indicates that the PLCP frame carries several MAC frames in an
aggregate burst.
Scrambler initialization (2 bits)
These two bits are used to seed the scrambler.
20/40 BW (1 bit)
If set to one, this bit indicates that a 40 MHz channel is used. When set to zero, it indicates a
20 MHz channel.
CRC (8 bits)
The CRC is used to protect the legacy signal field, and all the fields in the HT-SIG field before
the CRC.
Tail (6 bits)
The HT-SIG field is protected by a convolutional code, and as always, six bits are needed to
ramp down the convolutional coder.
High-Throughput training fields
Following the high-throughput headers are high-throughput short and long training fields. A single
short training field spans the entire operating channel. In 20 MHz channels, the bandwidth of the
high-throughput short training field (HT-STF) is 20 MHz. When the wider 40 MHz channels are used,
the HT-STF has a bandwidth of 40 MHz. The short training field fine-tunes the receivers in MIMO
operation.
When several spatial streams are transmitted over several chains, finer control of the amplification
applied to the incoming signal is important. Long training fields (HT-LTFs) are used to further tune
each receiver chains. One HT-LTF is used for each spatial stream. In basic MIMO mode, there is one
receiver chain for each spatial stream; in the advanced mode, there may be more receiver chains
than spatial streams.
Data, tail, and padding
Data bits are encoded according to modulation and coding methods that are defined by the highthroughput header. Like other OFDM physical layers, data is scrambled before transmission, using
the scrambler initialization bits in the high-throughput header. Following the data, there is a six-bit
tail that ramps down the convolutional code, and enough padding bits to make the data to be
transmitted equal to the symbol block size.
TGnSync PMD
The basic design of a TGnSync transceiver is shown in Figure 15-19, which depicts the design of a
beamforming transmitter, rather than a basic MIMO transmitter. An incoming scrambled frame is
handed to the forward-error correction coder, which is usually a convolutional coder. Coded bits from
the FEC coder are then sent to different spatial streams by the spatial parser, which is responsible for
dividing the unified bit stream into subsidiary streams for transmission. Each spatial stream is
punctured up to the desired rate. In the beamforming mode, the puncturing occurs for each spatial
stream and may occur at different rates. Basic transmitters must puncture every stream to the same
rate. (Logically, the puncturing in the basic MIMO transmitter can occur before the spatial parser.)
Each spatial stream now consists of a sequence of coded bits, ready for mapping on to OFDM carriers
by the interleaver. After the interleaver, each block of bits can be mapped on to a single symbol by
the constellation mapper. In the basic MIMO mode, each interleaved spatial stream is processed by a
single transmission chain; the advanced mode uses a spatial steering matrix to assign symbols to
any transmit chain. The spatial steering matrix shown in the figure could be replaced by a one-to-one
interface between the spatial stream processors and the transmit antennas for basic MIMO operation.
Each transmit chain takes its symbol sequence and modulates it on to the airwaves. There is no
mention in the specification of required channel rejection or sensitivity performance.
Figure 15-19. 2x2 TGnSync MIMO transceiver
Comparison and Conclusions
Both proposals are essentially MIMO evolutions of the 802.11a PHY. Both require support for a 2x2
mode, where both the sender and receiver have two transceivers active. However, it is likely that
most products that are based on the eventual 802.11n standard (which may or may not resemble
either of the proposals) will support at least some of the optional modes. It is likely that cost
constraints on client devices will restrict them to operating in a two-transceiver mode, while APs will
have more transceivers. Basic APs may have only two, while the most expensive enterprise-class APs
have three or four transceivers.
Table 15-5 shows the data rates for the two spatial stream modes for each proposal. Higher data
rates are possible with additional spatial streams, albeit at the extra cost of silicon. TGnSync's
proposed peak rates are higher, although at the cost of more aggressive coding. WWiSE's 135 Mbps
data rate is accomplished by using 64-QAM at R=5/6; TGnSync gets to 140 Mbps by using a 7/8 rate
code and cutting the guard interval to half. (Without the short guard interval, TGnSync's data rate is
only 126 Mbps.) The advanced beamforming mode uses the much larger 256-QAM constellation.
Though TGnSync's data rates are higher, I expect that the more aggressive coding will lead to
shorter range.
Table 15-5. Top speed for major 802.11n proposals (two spatial streams)
20MHz channels
40 MHz channels
135 Mbps
270 Mbps
Basic mode
140 Mbps (+3.7%)
315 Mbps (+16.7%)
Advanced beamforming mode
160 Mbps (+18.5%)
360 Mbps (+33.3%)
WWISE
TGnSync
Spectral usage is a major point of contention between the two groups. WWiSE has focused much
more heavily on improving MAC efficiency than on the data rate, even going so far as to argue that
using 40 MHz channels to improve the data rate before improving efficiency is a waste of scarce
unlicensed spectrum. While there may be merit to that point, the 40 MHz channelization approach
used by TGnSync has the advantage of being able to reclaim spectrum in the middle of a wider
channel. WWiSE merely doubles throughput in their 40 MHz mode, while TGnSync squeezes more
than double the capacity out of the wider channel. Both approaches have their drawbacks. TGnSync's
proposal would probably lead to chipsets that are always capable of 40 MHz operation, adding extra
cost and complexity, even though regulators may not allow them. In countries where 40 MHz
channels are allowed, the extra speed would be welcome. In areas where 40 MHz channels are only a
pipe dream of chipset manufacturers, the extra cost may not be welcome. On the other hand,
WWiSE's denial of the need for high-speed channels seems to be denying the five-fold leap in data
rates that occurs with every new 802.11 PHY.
For maximum speed, TGnSync requires closed-loop operation, which would be a major undertaking
to implement in silicon. Sounding frames must be used to measure the channel, and responses must
be collected to calibrate the radio channel. WWiSE uses only open loop operation, which is simpler to
implement. The WWiSE proposal also offers the ability to spread a single encoded stream across
multiple antennas without using closed-loop operation. If closed-loop operation were to be
problematic to implement in silicon, 802.11n could be delayed unacceptably.
Frame aggregation is an important part of meeting the larger goal set for the eventual 802.11n
standard. However, taking full advantage of aggregation opportunities requires more intelligent
queuing than is currently implemented. Whether 802.11n offers a huge increase in speed is likely to
depend a great deal on how well improved queuing algorithms are able to coalesce collections of
small packets into large aggregates. Neither proposal specifies queueing, so the performance boost
may vary between vendors.
Aggregation as designed by the protocol is a bit more intelligent in TGnSync, although this is only a
minor advantage. WWiSE's proposal only allows aggregation when the Address 1 field in the MAC
header is the same. In an infrastructure network, the Address 1 field is the BSSID. All frames from a
station to an AP can be aggregated, so the two proposals are identical in the upstream direction. In
the downstream direction, WWiSE must use a physical-layer frame burst to change directions. Each
new direction must have a new PLCP header. TGnSync can reduce overhead by using a multiplereceiver aggregate frame, and collecting responses from each receiver in the aggregate.
Powersaving modes in 802.11 have been neglected, and are not very sophisticated. TGnSync
attempts to come up with extended powersaving operations for some of the new MAC structures,
while WWiSE does not. This is likely due to the presence of equipment vendors that use chips in the
TGnSync consortium, while WWiSE is comprised only of chip vendors. While the trade-off between
high speed and battery life is application-specific and may not always make sense, it is always good
to see the standards bodies thinking ahead about problems that may occur.
Chapter 16. 802.11 Hardware
When writing a specification, it is important to leave some room for interpretation to allow
innovation. A standard which is too rigid will drain the life out of implementation, while a standard
that is too loose does not foster interoperability. This chapter is about what the standard does not
exactly specify. How does hardware implement the standard? Where does the standard allow
discretion in protocol implementation, and what does that mean for the administrator?
General Structure of an 802.11 Interface
Figure 16-1 shows a block diagram of a generic wireless LAN interface. It is not representative of any
particular manufacturer's product, but is intended to serve as a general guide to the discussion of
how cards are put together. Cards must implement the physical interface and the link-layer control
expected by the operating system.
Like any other system based on radio technology, wireless LAN interfaces have antennas. Most
802.11 interfaces have two antennas for antenna diversity to improve reception in the presence of
multipath interference. When a radio signal is received, the radio system selects the antenna with the
strongest signal, and runs with it. Antenna diversity alleviates multipath because one of the two
antennas should receive a signal if the weak signal is caused by self-interference of multiple signals.
Diversity cannot help if the signal weakness is caused by distance because both antennas will receive
equally weak signals. Many types of diversity are possible, but by far the most common
implementation is to perform diversity reception on received frames only. Most, if not all, products on
the market do not have any diversity for transmitted frames, and will always use a "primary"
antenna for transmissions.
Early antenna diversity implementations were helpful, but limited by lack of sophistication. Multipleinput/multiple-output (MIMO) technology promises higher data rates and better performance.
Detailed descriptions of MIMO are beyond this book. In essence, it improves the receiver's
performance by using both antennas rather than choosing one exclusively.
Figure 16-1. Generic wireless card structure
Antennas deliver the radio signal to a transceiver. The transceiver is sometimes called a radio
"modem" by analogy with the modems commonly used for telecommunications. Transceivers will use
amplifiers to boost the outgoing signal for distance, or the incoming signal for further processing.
Radio transceivers also downconvert the high-frequency signals into much more manageable signals
by extracting the data bits from the high-frequency carrier waves. Transceiver systems are usually
shielded to prevent the high frequencies in use from interfering with other system components,
which is why radio interfaces typically have a metal shield covering a good chunk of the interface
card's area.
After the transceiver, the next step in the chain is the baseband processor, which is the interface
between the digital and analog parts of the wireless LAN system. Bits come from the computer, and
are turned into radio waves for the antennas. Turning bits into radio waves is called modulation; the
reverse process is called demodulation. The baseband processor is responsible for handling the
complex spread-spectrum modulations, as well as physical carrier sensing. Generally speaking,
physical carrier sensing is integrated with demodulation. When radio energy is received in excess of a
threshold, the baseband processor attempts to demodulate it. Many different modulation techniques
exist for use with wireless LANs. One of the reasons why older 802.11b stations are blind to the
existence of 802.11g transmissions is that the older baseband processors used in 802.11b stations
are not capable of demodulating OFDM transmissions.
At the heart of the interface is the medium access controller (MAC), which is responsible for taking
incoming frames from the host operating system's network stack and deciding when to squirt the
frames out the antenna into the air. The MAC accepts frames from the operating system for delivery
through a system bus interface. Most wireless LAN interfaces are based on the CardBus standard;
some interfaces use Mini-PCI. On the other side of the MAC, it ties into the baseband processor to
send frames to the radio system.
MACs may work with several frames at once, and can use a small RAM buffer to store frames being
worked on. One of the most common reasons to work on a frame is to offload security functions from
the host operating system. Rather than requiring frames to have been encrypted by the main system
CPU before transmission, a MAC with security functions can accept frames from the driver that are to
be encrypted with a specified key. Frequently, such MACs implement a key cache, which can store
multiple keys. Drivers queue frames with a notation to encrypt with key 1 before transmission, and
the MAC offloads the encryption from the host system. Older MAC chips could perform RC4
encryption offload for WEP and could be retrofitted for TKIP, while the newest MAC chips are able to
perform AES encryption in hardware.
In addition to the RAM buffer for frames in process, most interfaces also have a small flash memory
to store firmware for the MAC. When the MAC is powered up, it retrieves and runs code from flash.
To implement a new security protocol like TKIP, put new firmware in the flash and reboot the MAC.
Firmware allows the use of relatively generic processors rather than hard-to-change applicationspecific integrated circuits (ASICs), which is a great help when working with protocols like 802.11
that may change quickly. Most 802.11 MAC chips are actually fairly generic microprocessors.
Some MACs implement a "real-time" section of the 802.11 MAC for items that require immediate
response. Rather than requiring the host operating system to respond to power-save polling
operations or send acknowledgments, these frames are often automatically generated by the MAC. In
other systems, the real-time functions are handled by the baseband processor. (Most modern
systems implement both the MAC and baseband in a single chip, so the distinction is no longer of
burning relevance.)
The block diagram of Figure 16-1 is a general guide only. To reduce cost and complexity, many
systems use a single chip for both the MAC and the baseband processor. Some solutions even
implement the radio transceiver on the same chip as the MAC and baseband controller. Interfaces
based on Atheros chipsets do not have flash memory for firmware because Atheros chips are given
code when the MAC is started by the driver. Firmware generally refers to code stored with the
hardware; Atheros devices can be programmed by system software.
Software-Defined Radios: A Digression
Two significant problems with 802.11 are the inefficiency of the MAC, which limits throughput, and
the rapid change in physical layer technology, which makes devices obsolete quickly. A natural
reaction to the inefficiency, especially from technically adept engineers, is to want to implement
specific optimizations or new protocol features on the hardware they already own.
With wireless LANs, customizing the behavior of the interface is generally not possible. The baseband
processor, or the functions of the baseband processor, are implemented directly in the integrated
circuit for speed, and cannot be changed. So-called application specific integrated circuits (ASICs)
have particular logic laid out in their circuitry, and cannot be used for anything other than what they
were designed for. Furthermore, wireless LAN interfaces use radio components optimized to run with
the frequencies allocated to wireless LAN usage by regulators.
However, future radios may be reprogrammable. Instead of using dedicated hardware for the
baseband processor, designers can use "programmable logic" devices, such as Field Programmable
Gate Arrays (FPGAs). Radio waves are analog waves that carry digital data. In many cases, radios
only need programmable signal processing, and can use digital signal processor (DSP) chips. No
matter what type of programmable logic is used, the end goal is the same. Rather than having a
modulation fixed in place during the design phase, radios based on programmable logic can change
modulation, coding, and bit rate on the fly with new software. The price of the flexibility is that
programmable logic devices are bigger, run slower, and consume more power (and therefore
generate more heat) than nonprogrammable devices.[*] By using programmable devices, the
behavior of the radio can be changed at will, simply by loading new software. New modulation types,
bit rates, or even new frequency bands can be implemented without changing hardware.[ ]
[*]
Programmable logic devices are often used in the design of application-specific devices. After the logic is designed and
thoroughly tested, the FPGA will be redesigned as an application-specific device. The resulting fixed-logic device is sometimes
called a "hard copy" of its programmable ancestor.
[
] It's intereting to note that the open source process has started to adopt programmable logic. OpenCores
(http://www.opencores.org) is the repository of hardware designs in the same pattern that SourceForge
(http://www.sourceforge.net) is for software.
Radio interfaces controlled completely by software (or programmable logic) are called softwaredefined radios or universal radios. In environments where the modulation and coding change rapidly
or frequently, they are extremely useful. One of the first major software-defined radios was the
SPEAKEasy project, launched by the U.S. military to replace 10 radio systems using different
modulations with a single software radio that could perform the same work as all 10 radio systems,
but in a much smaller package. Software-defined radios are subject to certification rules in the United
States that couple the radio software with the hardware, and require manufacturers to take steps to
avoid alteration of the software to operate outside the approved range. [ ]
[ ]
The FCC rules regarding the certification of software-defined radios are in Appendix B of FCC 01-264, which is available at
http://www.fcc.gov/Bureaus/Engineering_Technology/Orders/2001/fcc01264.pdf. The first software-defined radio, a GSM base
station developed by Vanu, was certified in November 2004.
Although software defined radios are an interesting development worth watching, the full
programmability will probably not be implemented in 802.11 devices. (Limited programmability is
used by both Broadcom and Atheros in their current chipsets.) Programmable logic is much more
expensive, and would price the resulting 802.11 devices out of the market. It may find a foothold for
use with 802.11 in a radio that implements several types of radio link, of which 802.11 is one.
A Few Words on 802.11 Hardware Implementations
At the time this book was written, there were four major chipsets used to build 802.11 vendors. In
alphabetical order, the established vendors are:
Atheros
Most interfaces that include 802.11a are based on Atheros chipsets. Many 802.11g devices are
based on Atheros chipsets as well.
Broadcom
Most of the non-Centrino built-in 802.11g interfaces available at the time this book was written
were based on the Broadcom 802.11g chipset. Apple's AirPort Extreme is based on Broadcom
silicon.
Conexant (Prism)
The Prism chipset line went through several corporate transitions before it wound up at
Conexant.
Intel (Centrino)
Many built-in wireless LAN interfaces in laptops are based on Intel's "Centrino" brand.
Technically, Centrino is a marketing term that refers to a whole set of Intel silicon, including
the system CPU. For example, the Intel/PRO 2200 card is the the Centrino 802.11g interface.
One of the more notable new chipset vendors is Airgo Networks, which is building a MIMO chipset
that underlies prestandard 802.11n products on the market.
It is often useful to know the chipset vendor behind the card. Most of the chipset vendors make
reference designs available to their customers. Reference designs typically include both hardware
and software. 802.11 card vendors can take a reference design, modify it slightly (or not at all),
arrange for a new label to be put on the manufactured device, and start selling it. Very few card
vendors make significant enhancements to the driver, and will simply repackage reference drivers.
Naturally, the track record of different vendors varies. Some make updated drivers available quickly,
others may not. Knowing who supplied the chipset in the card can make it easier to obtain either the
reference driver, or a more recent driver for another similar card using the same chipset. Naturally,
knowing the chipset is vital for picking out the correct driver for open souce operating systems.
Learning more about cards: FCC filings
802.11 interfaces are, in the language of regulatory agencies, intentional radiators. Rather than
simply complying with a standard that limits the amount of radio frequency energy given off, 802.11
interfaces are designed to spew radio waves into the air. Intentional radiators are tested for
compliance with every country's regulatory requirements, which generates a great deal of
paperwork.
In the United States, radio devices are regulated by the Federal Communications Commission (FCC),
and radio transmission devices are tested for compliance with FCC rules. Before a device can legally
be sold, it must be assigned an identification number. Take a look at your favorite card and look for
the FCC ID.
An FCC ID is composed of two parts: the grantee code, which is the first three letters, and the
product code, which may be up to 14 letters and numbers. Grantee codes are unique to an
organization. (The grantee code may be separated by a space, dash, or not at all.) For example, the
FCC ID for a Lucent Gold card is IMR-WLPCE24H. IMR is Lucent's grantee code, while WLPCE24H is
specific to the Gold card.
As part of the testing process, manufacturers must submit test reports, product photographs, and
other documentation, which becomes part of the public record. To look up information on your
favorite card, go to the search engine maintained by the FCC's Office of Engineering Technology at
http://www.fcc.gov/oet/fccid/.
FCC identifiers can be useful if you are trying to figure out what driver to use with a particular card.
Most products have internal photographs available, which may enable you to determine whose
chipset is in use. Say, for example, you had a Proxim Gold a/b/g combo card and wanted to know
what chipset it used. By searching the FCC database on its ID (HZB-8460), you could obtain internal
photographs of the device, which clearly show the presence of Atheros chips on the system board.
Implementation-Specific Behavior
802.11 is not a rigorous standard. Several components of the standard are relatively loose and leave
a great deal up to the particular implementation. Most implementations are also relatively young, and
may behave in a nondeterministic fashion. I have participated in tests that made use of several
identical computers, all imaged from the same software distribution and using the exact same wireles
LAN hardware and driver revision. Even though there were several computers in the same location
with identical configuration, behavior was significantly different.
Rebooting Interface Cards
802.11 is a complex protocol with many options, and running the newest protocols exposes the
newest bugs. 802.11 interfaces use relatively general-purpose microprocessors running software. As
with a great deal of software, cards that are in a strange state may be helped by "rebooting" to clear
any protocol state stored in the MAC processor. External cards can be rebooted by removing and reinserting them; internal cards must be rebooted by power cycling them through the system software.
It is not sufficient to unload and reload drivers, since the object is to clear all state in the wireless
LAN interface.
Restarting a card may be required to clear state that is preventing successful operation. As a first
step, restart the card when:
The client system is associated, but cannot send or receive traffic. If the network is encrypted,
the problem is often a lack of synchronization of cryptographic keys. This problem is often
exacerbated by roaming because every change between APs results in the transmission of new
keys.
No scan list can be built. If you are sure that there is a network within range but it will not show
up in the client utility, the card may be in a state where it is unable to supply a scan list.
Authentication/association failures occur in rapid succession. If the state of the client system
software prevents a successful connection but the network is on a "preferred" list for
connection, the attempt will be retried.
Scanning and Roaming
Every card behaves differently when searching for a network to attach to, and in how it decides to
move between APs. 802.11 places no constraints on how a client device makes its decision on how to
move between APs, and does not allow for any straightforward way for the AP to influence the
decision. Most client systems use signal strength or quality as the primary metric, and will attempt to
communicate with the strongest AP signal.
Most cards monitor the signal-to-noise ratio of received frames, as well as the data rate in use, to
determine when to roam to a new AP. When the signal-to-noise ratio is low at a slow data rate, the
client system begins to look for another AP. Many clients put off moving as long as possible, in part
because the process of looking for a new AP requires tuning to other channels and may interrupt
communications in process. Client stickiness is sometimes referred to as the bug light syndrome.
Once a client has attached to an AP, it hangs on for dear life, like a bug drawn to a bug zapper. Even
if the client moves a great distance from the AP with a consequent drop in signal strength, most
clients do not begin the roaming process until the signal is almost lost.
Roaming in 802.11 is entirely driven by client decisions. Where to send the Association Request
frames is entirely in the hands of the client system's driver and firmware, and is not constrained by
the 802.11 specification in any way. It would be 802.11-compliant, though awful, to connect to the
AP with the weakest signal! (An unfortunate corollary is that driver updates to fix bugs may alter the
roaming behavior of client systems in undesirable ways.) Access points do not have protocol
operations that can influence where clients attach to, and whether they will move or not.
Implementing better roaming technology is a major task for 802.11 as time-critical streaming
applications begin to use 802.11.
To borrow a phrase from Milton Friedman, roaming is always and everywhere a
client phenomenon.
Rate Selection
802.11 lays out basic ground rules for how multirate support needs to work, but it leaves the rate
selection algorithm up to the software running on the interface. Generally speaking, an interface tries
to transmit at higher speeds several times before downgrading to lower speeds. Part of that is simply
common sense. In the time it takes to transmit a frame with a 1,500-byte payload at 1 Mbps, it
would be possible to transmit the same frame 8 times at 11 Mbps, or over 20 times on an 802.11g
network running at 54 Mbps with protection enabled. (Without protection, the multiplier is 40!) If the
frame was corrupted by a one-time event, it makes sense to retry a few times before accepting the
more drastic penalty of lowering the data rate.
Step-down algorithms are generally similar. After trying some number of times to transmit a frame,
it falls to a lower data rate. Most cards step down one rate at a time until an acknowledgment is
received, though there is no requirement for them to do so. It would be a valid rate selection
algorithm to slow down to the minimum data rate at the first sign of trouble. Step-up algorithms
work the same way in reverse. When "several" frames are received with a much higher signal-tonoise ratio than is required for the current rate, the interface may consider stepping up to the next
highest rate.
Reading the Specification Sheet
Early testing of 802.11 devices focused on range and throughput because there was little else that
could be tested. In some environments, range may be an important factor, and one that can be
determined in part by reading the specification sheet for a card.
Range is largely a function of receiver sensitivity, which is a measurement of the weakest signal that
a receiver can correctly translate into data. Better sensitivity measurements result in improved
range. (Improving sensitivity can also help other performance factors, too, but range is the easiest
one to discuss.) Most vendors have focused on improving performance, and the resulting
improvements (Atheros' XR, and Broadcom's BroadRange) have been announced with great fanfare.
Not all vendors publish complete specification sheets. Cisco discloses a great deal of information,
with receiver sensitivities disclosed at each data rate in all frequency bands supported. (The card's 5
GHz performance has a slight dependence on the frequency.) Many vendors simply report the
supported data rates without any indication of sensitivity.
Sensitivity Comparison
As an example, compare the sensitivity of some common 802.11b cards. Sensitivity is defined by
each 802.11 PHY. For the direct sequence rates, it is defined as the received power at which the
input frame error rate is 8%, for 1024 byte frames. The standards require that the sensitivity be -76
dBm or better for 11 Mbps, and -80 dBm for 2 Mbps.[*] Lower sensitivity is better because it means
the card can receive weaker signals than required.
[*]
This corresponds roughly to a bit error rate of .001% (10-5).
The sensitivities reported in Table 16-1 were taken from data sheets and user manuals for four wellknown 802.11b cards and a new a/b/g card. The Cisco Aironet 350 had a reputation for pulling in
weak signals well, which is entirely justified by the data. At 11 Mbps, it was sensitive to a signal half
as strong as the Orinoco Gold card, and nearly a quarter of the signal required by the Microsoft card.
However, the march of technology has improved sensitivities at higher bit rates. All the oldergeneration cards are less sensitive than the Atheros-based Cisco tri-mode card currently on the
market.
Table 16-1. Sensitivity (in dBm) for various cards
Card
11 Mbps
5.5 Mbps
2 Mbps 1 Mbps
Cisco 350
-85
-89
-91
-94
Orinoco Gold (Hermes)
-82
-87
-91
-94
Linksys WPC11 (Prism)
-82
-85
-89
-91
Microsoft MN-520
-80
-83
-83
-83
Cisco CB-21 (a/b/g); 802.11b performance only
-90
-92
-93
-94
Delay Spread
When radio waves bounce off objects, several echos of the wave will converge on the receiver. The
difference between the first wave's arrival and the last arrival is the delay spread. Receivers can pick
through the noise to find the signal, but only if the delay spread is not excessive. Some vendors also
quote the maximum delay spread on their data sheets. Table 16-2 reports the delay spread for three
of the cards listed above.
Table 16-2. Delay spread (in ns) for various cards
Card
11 Mbps
5.5 Mbps
2 Mbps 1 Mbps
Cisco 350
140
300
400
500
Orinoco Gold (Hermes)
65
225
400
500
Cisco CB-21 (a/b/g); 802.11b performance only
130
200
300
350
Cards rated for higher delay spreads are capable of dealing with worse multipath interference. Again,
the Cisco Aironet 350 was an extremely capable card for its day, capable of dealing with over twice
the time-smearing as the Hermes-based card.
Chapter 17. Using 802.11 on Windows
Whether you've made it to this point by skipping Chapters 3-16, or whether you've read all the
theory, we're now going to get our hands dirty and start installing equipment.
Development of 802.11 management interfaces has followed the familiar progression of Windows
applications. In the beginning, there was a great deal of variation between individual vendors. As
popularity grew, Microsoft integrated 802.11 configuration into the operating system, subsuming
vendor-specific management tools into an overall framework.
From the standpoint of practical system and network administration, working with 802.11 is similar
to working with Ethernet. Installing 802.11 drivers is nearly identical to installing Ethernet drivers,
and the network interfaces behave almost exactly like Ethernet interfaces. 802.11 interfaces cause
an ARP cache to be brought into existence, and other software may even perceive the wireless
interface as an Ethernet interface. Unlike many Ethernet drivers, however, 802.11 drivers can have a
number of advanced knobs and features that reflect the additional management features presented
in Chapter 8.
This chapter discusses Windows configuration of wireless cards on both Windows XP and Windows
2000. I strongly advise using Windows XP for wireless-enabled machines because it is generally
easier to use and has substantial additional support for new protocols. Third-party supplicants
generally disable the built-in supplicant. Occasionally, however, a driver will refuse to work with a
security system other than Microsoft security stack.
Windows XP
Windows XP continues the long-time Windows practice of incorporating functionality that was
previously only supplied by third parties into the operating system itself. Rather than relying on each
802.11 card manufacturer to supply a configuration utility, Microsoft has built a standard
configuration utility, called Windows Zero Configuration (also called ZeroConfig, ZeroConf, or WZC),
which works with most cards. Card manufacturers must still provide driver software, but the
configuration can be handled through Windows screens rather than card-specific programs. This
section discusses ZeroConfig because it presents a card-independent view of configuration, which
makes it popular with network administrators and help desks.
Card Installation
Before starting the process, it generally pays to get the latest drivers for your card from the
manufacturer's web site. Other network technologies are far more mature than 802.11, and do not
require frequent driver updates. Unfortunately, rapid innovation in wireless is accompanied by rapid
driver changes.
For major vendors, this is usually found in a "support" or "downloads" section. Many smaller vendors
are rebadging reference designs, and may not have drivers available. In such cases, it is often
possible to use the reference driver from the chipset vendor.
Many drivers ship with an installer, which I generally prefer not to run. Several cards have licensed
third-party 802.1X stacks, which can interfere with the operation of ZeroConf. It is usually best to
obtain the latest drivers from the card vendor's web site before installing the card.
Third-party 802.1X stacks and the driver update process
Several third-party 802.1X stacks exist. On Windows, they are generally implemented as a shim in
the network stack that intercepts EAPOL frames between the hardware and the network protocol.
Many card vendors, frustrated with the progress of the built-in Microsoft supplicant, have turned to
third-party stacks to enable 802.1X. Many third-party stacks are incompatible with ZeroConfig.
Rather than use the ZeroConfig screens to set parameters, third-party stacks often require an
additional software configuration tool. They may also require additional supplicant software to
support advanced protocols.
To avoid using the bundled 802.1X stack, update the driver without running the installer from the
vendor. The best practice is to get the driver from the vendor web site, unzip it, and then use the
update driver panel of the device manager to roll the new version of the driver into your operating
system configuration. Figure 17-1 illustrates how to upgrade only the driver. By going to the Driver
tab of the hardware properties, you can access the Update Driver button.
Two tricks may be used to avoid installing the third-party 802.1X stacks. Many executable driver
installation files are self-extracting ZIP packages, and can be opened by an unzip program. Extract
the drivers to a temporary location, and update from a
Figure 17-1. Updating the driver without the baggage
file. An alternative is to run the software installer, and immediately run the uninstaller. Most
uninstallers will remove configuration tools and third-party shims while leaving the new driver in
place.
Cisco client software
Cisco cards have one additional pitfall. Protected EAP (PEAP) is still a developing standard. Although
written by representatives from both Cisco and Microsoft, the two companies' implementations are
not compatible. Most of the industry has shipped PEAP implementations that will autodetect which
version is in use.[*] Even Cisco's authentication server, CiscoSecure ACS, has a configuration option
to be compatible with the Microsoft PEAP implementation.
[*]
Both the Funk and Meetinghouse clients do this, as does the Radiator RADIUS server.
Microsoft's PEAP implementation supports the use of EAP-MS-CHAP-V2 and EAP-TLS as inner
authentication protocols; Cisco's PEAP implementation supports EAP-SIM and EAP-GTC. Most
Windows shops will be seeking to use an existing user database, such as an Active Directory, which is
possible only with EAP-MS-CHAP-V2.
As part of the bundled software set, Cisco ships replacement (Cisco) PEAP software that overwrites
the standard (Microsoft) PEAP implementation on the system. To use Microsoft PEAP, you must
remove the Cisco PEAP drivers and replace them with Microsoft PEAP. There is only one set of PEAP
software for the entire system, so installing Cisco PEAP will affect all cards on the system, not just
Cisco cards. See Table 17-1.
Table 17-1. Inner EAP methods
EAP method
Cisco or
Microsoft
PEAP?
Authentication credentials
Notes
EAP-MS-CHAPV2
Microsoft
Used with Windows domain
Shared password (or MD4 hash
authentication; easy to plug in to
of the password)
Active Directory
EAP-TLS
Microsoft
User certificate
Generally not used
Cisco
Subscriber Identity Module
(SIM) card
Based on GSM mobile telephone
authentication; not yet widely
used
Cleartext authentication string
passed through encrypted
tunnels
Can be used for token cards as
well as a generic method for static
passwords
EAP-SIM
EAP-GTC
(Generic Token Cisco
Card)
Many laptop vendors offer 802.11 cards preinstalled into the system, with software preloaded and
configured. When the card in question is a Cisco card, chances are that the laptop vendor has
installed Cisco PEAP and your organization will need to adjust its system build process to restore
Microsoft PEAP. To restore Microsoft PEAP, you must reinstall the latest service pack to overwrite the
Cisco PEAP drivers with the Microsoft PEAP drivers.
Choosing a Network
When Windows starts up, the ZeroConfig system will attempt to find any networks in the area.
ZeroConfig may automatically connect to networks which it already has a configuration for. If no
networks have yet been configured, the wireless interface will produce the bubble shown in Figure
17-2.
Figure 17-2. Prompt that WLANs are in the neighborhood
By clicking on the box, or by right clicking on the wireless interface and choosing "View Available
Wireless Networks", you can get to the list of detected networks (Figure 17-3). By choosing to
Connect to any one of them, Windows will attempt to automatically configure the appropriate
settings. However, it has no way of selecting the appropriate EAP method, and therefore is often
unable to automatically configure encrypted networks. Attaching to an unencrypted network is easy,
but you will be prompted to confirm connecting to a network that provides no security.
Figure 17-3. Viewing available wireless LANs
Configuring Security Parameters and 802.1X
In many cases, it is necessary to manually configure the security properties of a network. Windows
can automatically default settings for encryption and authentication type, but may still be wrong.
802.1X settings are always defaulted to using EAP-TLS, even though EAP-TLS is not commonly used.
To change security parameters, go to "Change Advanced Setttings" to get the interface properties,
and hit the Wireless Networks tab (Figure 17-4).
To add a network, you can select it from the Preferred Networks pane and choose Properties, or you
may add it manually by specifying a network name, as in Figure 17-5. If you are manually adding a
network, you must configure the correct properties from scratch. If Windows has automatically
configured the network, it will pick the strongest protocols available.
Figure 17-4. Wireless Networks tab
One of the positive attributes of ZeroConf is that it can maintain properties for several different
wireless LANs, and connect to networks for which you have defined settings appropriately.
The first major choice to make is the type of network authentication you will use. In the figure, I
show WPA. However, there are four options:
Open authentication.
This is used for networks that perform no authentication, as well as networks that are using
legacy dynamic WEP solutions. Selecting it means that the initial link layer authentication takes
place using the open authentication method described in Chapter 8. Any network implementing
open authentication may require 802.1X on top of it, though it does not have to.
Shared authentication.
This option is used for networks that do shared-key WEP authentication. It will require you to
enter a key in the "Network Key" fields. Shared-key WEP authentication is not very strong, and
does not offer serious protection. The only reason to use this option is to connect to a network
of APs that requires it.
Figure 17-5. Association properties
WPA authentication.
This is short for WPA Enterprise. It uses 802.1X for user authentication, and uses the
authenticated keying mode defined in 802.11i. This option is only available if WPA software is
installed and the driver has support for WPA.
WPA-PSK authentication.
This is short for WPA Personal. Rather than derive the master secret, this method uses a
preshared master key. Link encryption keys are derived from the preshared key and random
values exchanged by both the client and the AP.
After choosing the authentication type, you will need to select the encryption used on the link.
Disabled.
This is used on open networks that do not apply link layer encryption. Many 802.11 hot spots
currently fall into this category, although most should offer stronger encryption in the future.
WEP.
This includes WEP using both manual and dynamic keys. To use dynamic WEP, check the box
that says "The key is provided to me automatically". Manual WEP keys can be configured by
unchecking the box and entering the key.
TKIP.
This option is used for TKIP, which is the default value for WPA networks.
AES.
Windows does not refer to CCMP, but instead labels it AES in the ZeroConf configuration.[*]
This is believed to be the strongest encryption available for use on wireless networks.
[*]
This is probably due to intellectual property concerns that caused the AES-based encryption algorithm
to be changed during the 802.11i draft process.
The dialog box offers the option to enter a key, or use automatically provided keys. If 802.1X is in
use, the key will be provided automatically and the checkbox will be cleared. Networks that use WPA
authentication must use automatic keys; networks built on WPA-PSK must have the preshared key
entered for them. Table 17-2 summarizes the encryption and authentication options that are
supported.
Table 17-2. Summary of encryption and authentication methods
Authentication framework
Supported encryption
Authentication methods
Open
Disabled (no WEP)
802.1X optional
WEP (key specified)
WEP (automatic key)
WPA
WEP (automatic key)
802.1X required
TKIP
AES (supported cards only)
WPA-PSK
WEP (automatic key)
802.1X not used
TKIP
AES (supported cards only)
After configuring the association properties, configure the authentication properties. The association
properties tab, shown in Figure 17-6, is the screen on which the EAP method is selected. The figure
shows a machine which has three options: PEAP, EAP-TLS ("Smart card or other Certificate"), and
TTLS ("SecureW2"). Clicking on the properties button will configure the selected EAP method.
After the EAP method, there is a checkbox to "Authenticate as computer when computer information
is available". Checking this box allows a system to authenticate to the network using machine
credentials, which allows Windows domain housekeeping to occur before user authentication.
Operation of this process is discussed in a later section of this chapter.
Figure 17-6. Authentication properties
Configuring EAP Methods
On the authentication tab, you select an EAP method, but configuration is handled by clicking on the
Properties button. Not all the methods displayed in the authentication properties tab are supplied by
Microsoft. Third-party software can add EAP types to the drop-down list. Text in the EAP method box
is one of the ways that you can determine whether Microsoft or Cisco PEAP is in use. Microsoft PEAP,
which is technically PEAP version 0, displays as "Protected EAP (PEAP)", while Cisco PEAP displays
simply as "PEAP".
EAP-TLS
Figure 17-7 shows the configuration screen that is displayed when EAP-TLS is selected. The top box
specifies whether the client certificate used to authenticate the user is stored on a smart card, or in
the certificate stores that Windows maintains. Large organizations issue smart cards to employees,
sometimes embedded in employee identification card. Most midsized and smaller companies will opt
instead to issue certificates stored on each machine.
Server validation is a key to building a secure network. The Microsoft supplicant allows two different
levels of authentication. When the authentication server presents a certificate to the client, the client
can validate that against the list of trusted certification authorities. If a self-signed certificate is in
use, you must uncheck the validation box. The supplicant will trust any certificate it received, which
opens a huge potential route of attack because the supplicant will trust any AP with any certificate.
As an optional additional safeguard, the name listed in the server certificate can be validated to
ensure that it comes from a known DNS domain.
Figure 17-7. EAP-TLS configuration screen
PEAP version 0
Microsoft PEAP configuration, shown in Figure 17-8, is similar to EAP-TLS. Server validation is
identical, and should be enabled. The major configuration is the inner authentication method to be
used inside the protected tunnel; see Table 17-1 for a brief discussion of the various options.
Microsoft's implementation allows the list to contain any EAP authentication plug-in. By default, the
list contains EAP-TLS and EAP-MSCHAP-V2. The latter is more common because it enables easy
integration with Active Directory or NT Domain user accounts. Each inner authentication method can
have its own subconfiguration. With EAP-MSCHAP-V2, the inner method configuration has only one
option. When the box is checked, the Windows logon credentials are automatically submitted to the
network. The checkbox is sometimes called the PEAP single sign-on because the supplicant
automatically submits logon credentials without prompting the end user.
Figure 17-8. PEAP version 0 configuration
Clearing credentials from the registry
When single sign-on is not enabled, user passwords will be saved to the registry. When a user
connects to a particular network name for the first time, the system will prompt the user for a
username, password, and Windows domain. If authentication is successful, the credentials are saved
to the Registry for future use. Whenever the user chooses to connect to the same network in the
future, Windows will automatically submit the previously successful credentials.
As a user-friendly feature, credential caching leaves something to be desired, in large part because
there is not a convenient way to clear out user credentials that are no longer useful. Many wireless
networks plug in to existing authentication sources, and most large-scale networks enforce regular
password change policies. When an enforced password change invalidates the cached credentials in
the Registry, the user must clear out the registry to change the password.
User credentials are stored in the Registry under HKCU\Software\Microsoft\EAPOL\UserEapInfo, as
shown in Figure 17-9. Each network interface configured to do PEAP authentication without singlesign on will be represented by a long identifier consisting of a big string of letters and numbers.
Figure 17-9. Stored user credentials in the Registry
Each network card is given an identifier. In Figure 17-9, the interface with the identifier {768EB04A989E-4C89-9B85-93DF52F5EDE6} has three cached network passwords. To remove the cached
password for the network, delete its corresponding registry key.
To find out which network interface the string corresponds to, go to
HKLM\Software\Microsoft\WindowsNT\CurrentVersion\NetworkCards and check out keys below that
value. Figure 17-10 shows the mapping of the registry string to an "ORiNOCO 802.11abg ComboCard
Gold."
Figure 17-10. Cross-referencing the interface to the text name in the
Registry
Storing the user password in the registry decreases the usability of the Windows supplicant with a
user-entered password. Microsoft representatives have defended password caching as a user
convenience feature that prevents the need to repeatedly re-enter the password, and pointed out the
benefit of using an existing Active Directory user account with automatic use of Windows credentials.
Other supplicants, however, store the user password in a more easily accessible place.
The Microsoft supplicant on Windows CE does not cache credentials. Users are prompted for
credentials every time the supplicant reauthenticates. Most Windows CE devices do not have
keyboards, and entering credentials quickly can be quite difficult.
SecureW2: TTLS with ZeroConfig
Third-party supplicants present network administrators with a difficult choice. Users are often
comfortable with the Windows GUI for configuring wireless interfaces, but TTLS support is generally
only available by using a third-party client that disables ZeroConfig. Fortunately, there is a solution.
ZeroConfig has a programming interface that enables additional EAP support through "plug-ins" that
implement a particular authentication method. SecureW2 is a TTLS plug-in for Windows ZeroConfig.
It is an open source project, released under the GPL. Source code and binary packages are available
from http://www.securew2.com.
As a plug-in, SecureW2 configuration is accessed through the same screens as EAP-TLS and PEAP.
When the Properties button is first clicked, SecureW2 displays its main configuration screen (Figure
17-11). SecureW2 configuration data is bundled into a profile. Each wireless network stored by
ZeroConfig is associated with a SecureW2 profile. Different networks may each be associated with a
single SecureW2 profile. SecureW2 installers can be built with an administrative packaging tool that
allows profiles to be included in the installer.
Figure 17-11. SecureW2 main screen
Four tabs are used to define the profile. The Connection tab, shown in Figure 17-12, is used to
specify the first-stage identity used by TTLS. By unchecking the box, the user account specified on
the fourth tab will be used in the first stage. By default, the outer identity is anonymous.
Figure 17-12. SecureW2 profile creation
Server authentication is configured with the second tab, the Certificates tab shown in Figure 17-13.
Rather than presenting a long list of certificate checkboxes to accept, you must add each certificate
authority manually. By clicking the Add CA button, the list will pop up. In XP Service Pack 2, the list
presented is pulled from the computer's certificate store. Just as with the Microsoft supplicant, it is
possible to validate the server name in the certificate.
With the outer username and network authentication configured, it is possible to finally configure the
user account for the inner authentication. Authentication type is configured with the "Select
Authentication" tab drop-down (see Figure 17-14). Much of the impetus for using TTLS is that
cleartext password authentication is required; in such cases, the Authentication Method will be set to
PAP. The user account is configured on the fourth tab.
Figure 17-13. SecureW2 certificate configuration
Figure 17-14. SecureW2 authentication method and user account screens
WPA Configuration and Installation
Wi-Fi Protected Access (WPA) is the typical security baseline used in most organizations. WPA is built
into Windows XP Service Pack 2, and available as a patch for XP Service Pack 1. Enabling WPA is a
matter of selecting either WPA or WPA-PSK as the authentication method in the association
properties of the card. Microsoft does not implement WPA for other operating systems; use of WPA
on other operating systems requires a third-party supplicant.
To use WPA, the entire system must support it. In addition to OS-level support, the hardware and its
driver must be capable of WPA support. Drivers will prevent unsupported options from being
displayed. On a Dell TrueMobile 1150 card, the option for AES does not appear because that device
only has RC4 encryption built-in. If WPA is not working as expected, ensure that the latest driver is
installed before contacting your vendor's support organization.
Windows 2000
Windows 2000 is still widely used. Many older machines have not been upgraded to Windows XP to
save money, and Windows 2000 remains a serviceable operating system. Wireless configuration on
Windows 2000 is significantly more complex than on Windows XP, in large part because it lacks solid
integration between the selection of a wireless network and the corresponding security configuration.
Windows 2000 did not ship with 802.1X support from the start. It was initially added as a patch on
top of Service Pack 3,[*] and was later integrated into Service Pack 4. Microsoft has not ported WPA
functionality to Windows 2000, although a WPA client is available from the Wireless Security
Corporation (http://www.wirelesssecuritycorp.com). Many observers feel that 802.1X support on
Windows 2000 is not a priority for Microsoft, and its inclusion into recent service packs is an
illustration of increased difficulty in persuading users to upgrade.
[*]
See Microsoft knowledge base article 313664 for the patch.
Although 802.1X configuration has been integrated into the driver layer, Windows 2000 still depends
on a card utility to configure which network the system will attach to. The separation can be
particularly problematic for users who travel between an encrypted networks and unencrypted
networks. Although using the card utility to switch networks is straightforward, it is usually necessary
to manually enable or disable security. Windows 2000 may present network administrators with a
difficult choice. If the software configuration tool bundles a third-party 802.1X stack, extra
administration work must be done to separate the two.
Dynamic WEP Configuration
The Wireless Configuration Service on Windows 2000 only supports dynamic WEP for encryption.
TKIP support is only possible by using a third-party supplicant. To configure dynamic WEP, set up the
card's utility for use with manual WEP key. As far as the card utility is concerned, a manual WEP key
is in use. Frames are dispatched by the driver to the card, to be encrypted by one of the keys stored
in the card's key cache. The Wireless Configuration process, however, will push new keys into the
card as required by the network's security policy.
The manual WEP key need not be configured anywhere else on the network. It must only be the
correct length. For networks using 128-bit WEP, the key should be entered as 26 hexadecimal digits,
such as 12345678901234567890123456. This dummy key is never used, since it is replaced by the
dynamically derived key after a successful 802.1X authentication.
In my experience, the Wireless Configuration Service on Windows 2000 is not as reliable as the
process on Windows XP. Several bugs have caused the service to fail after a successful
authentication. Interestingly enough, the symptom of this type of failure is that the connection will be
keyed succesfully, but traffic will be disrupted at the first reauthentication period. With no software
running to process 802.1X frames, any attempted reauthentications or re-key operations will fail.
Windows Computer Authentication
When the Windows authentication subsystem was designed, it used a network connection to send
user credentials to the domain controller for validation. When the network subsystem started, it
would obtain a network address, if necessary for the protocol in use, and contact the domain
controller. Besides validating user credentials, the domain offered several other services. Network
administrators could define domain policies to control the behavior of any system in the domain, or
login scripts to customize the user environment as part of the login process.
In the wired world, domain services are not a problem. Users attach to the network, and the system
itself starts sending packets. When the Windows startup process was designed, there was no way of
authenticating wired network connections. In the wireless world, though, where users must
authenticate to active a wireless connection, network authentication can be a bit of a chicken and egg
problem. Users can certainly authenticate to the wireless network once they supply credentials to the
login box, but how can those credentials be validated without a network connection to send packets
to the domain controller? Windows NT, 2000, and XP provide a partial solution in terms of credential
caching. Once a user has successfully logged in to a computer that is a member of a Windows
domain, credentials are cached for the future.
Credential caching is only part of the solution because it depends on logging in to the computer on a
wired connection first. Microsoft has developed a much better solution called computer authentication
or machine authentication. When the system first starts up, the computer authenticates to the
wireless network as itself. With the wireless network up and operational, the computer can then
download any required information from the domain and validate the credentials of any user against
the domain controller. Users who have never logged in to the system and have no cached credentials
can log in as well.
A fair amount of functionality depends on having a network connection early in the boot process. In
addition to domain services, drive mappings are attempted early enough that they may fail if the
computer is not authenticated. In most cases, Windows computer authentication should be
considered mandatory for a smooth-functioning network.
To set up Windows computer authentication, you must have an Active Directory back end. Computers
must have accounts defined in the Active Directory, and they must be granted Dial-In permission.
How It Works
Computer authentication adds an extra 802.1X authentication process to the boot process. First the
computer authenticates as itself. After obtaining user credentials from the login box, a second 802.1X
transaction is run to authenticate as the user. The process is shown in Figure 17-15.
The figure identifies several steps in the process. It starts when the computer comes up and starts its
networking system. It begins an 802.1X authentication as the computer in parallel to other system
start-up tasks.
1. The machine authentication is started. It may be started by an EAPOL-Start from the
supplicant, or by a Request/Identity frame from the authenticator.
2.
1.
2. The network collects the identity of the machine. In the first authentication, the "user" identity
is of the form host/ComputerName.ActiveDirectoryDomain, where both the ComputerName and the
ActiveDirectoryDomain are available from the system properties.
3. The computer authenticates against its computer account on the RADIUS server, or in the backend database behind the RADIUS server. This process may take several round-trips because of
the need to exchange certificates, establish cryptographic keys, and so forth.
Computer authentication uses the same EAP method as user authentication. If EAP-TLS is
employed for user authentication, the computer must also have its own certificate. When PEAP
is configured for user authentication, the computer uses EAP-MSCHAP-V2 as its inner method.
The "password" used in the inner authentication is generated when the computer joins the
domain, and is not available to any other software.
4. When the authentication succeeds, the computer is attached to the network. It
receives an EAP-Success frame from the authenticator, and on wireless LANs, EAPOLKey frames to provide keys for the connection.
When authentication completes, the computer will be attached to the network, and can send
and receive traffic. By sending a DHCP request, the computer can join the network and locate a
nearby domain controller by using NetBIOS over
Figure 17-15. Startup process with Windows computer authentication
TCP/IP. Before user authentication, the computer can establish a relationship with the domain
controller. The user presses Control-Alt-Delete to begin the login process. The system uses its
connection to the domain controller to load the user login script, configure Windows domain
policies, and perform other login tasks.
5. When the user desktop is about to start, the second authentication starts. The end user
authentication must be triggered by the operating system. Typically, the machine authentication
established at the start of the session is only used for a few minutes until the user
authentication begins.
6. The network requests the identity of the end user. On Windows networks, this is typically of the
form DOMAIN\user, where the two components are the Windows NT-style domain name, and
the user account.
7. The supplicant authenticates to the network with credentials from the user. With the Windows
supplicant, the same EAP plug-in must be used for both the computer authentication and the
8.
7.
user authentication. Like the previous step, it may require multiple round-trips to establish a
secure cryptographic channel.
8. The user authentication succeeds, and keys are provided for the connection. These keys will be
different from the computer keys because they are derived from a different TLS session.
Computer authentications can be treated separately from user authentications. Authorization of the
two authentications may occur separately. For example, in a network that performs dynamic VLAN
assignment, it may be possible to assign the computer account to a different VLAN from the user
account. Early versions of the Microsoft supplicant did not trigger DHCP requests upon successful
authentication, which prevented the user from sending and receiving traffic. Patches are available to
fix this problem.
Chapter 18. 802.11 on the Macintosh
Apple Computer has been a key player in establishing the market for 802.11 equipment. Most
companies in the 802.11 market saw their contributions in terms of standards committee activity and
technology development. Apple contributed by distilling complex technology into an easy-to-use form
factor and applying its mass- marketing expertise.
In 1999, 802.11 was a promising technology that had demonstrated its value in a few narrow
markets. 802.11 interfaces cost around $300, and access points were around $1,000. Apple saw the
promise in the technology and moved aggressively, releasing $300 access points and $99 interfaces.
With a new competitor suddenly pricing the gear at a third of the prevailing price, other vendors were
forced to drop prices dramatically, and the market took off. Prices have been dropping ever since.
Apple's cards are branded with the name AirPort. AirPort refers to the first-generation 802.11b
network interfaces, while AirPort Extreme is used to refer to newer 802.11g-based hardware. (Due to
a focus on small offices and home offices, Apple does not sell 802.11a hardware.) This chapter
discusses only the AirPort Extreme hardware, although the differences in configuration and
management are vanishingly small. In addition to easy configuration of the wireless interface, the
Apple 802.1X supplicant is the easiest to configure. The 802.1X supplicant was included for the first
time in OS X 10.3, better known to most of the world by its code name of Panther.
The AirPort Extreme Card
Apple offers tightly integrated systems because the hardware and the software are designed in
tandem. Unlike the chaotic IBM-compatible world, with Apple one company is responsible for both
the hardware and software, and it shows. You can install the hardware and software and connect to
an existing network in only a few minutes. If an AirPort card is plugged in during system installation,
this can be done as part of the initial configuration.
Software Installation
Drivers for the AirPort are included in OS 9.1 and later, so there is no need to download and install
drivers. If the AirPort card is installed before the system first boots, the first-time boot configuration
utility will allow you to configure the AirPort interface out of the box by selecting a network name and
choosing how to configure TCP/IP. For a network that uses DHCP, the configuration instructions are
only a few screens long.
AirPort cards added after the system first boots can be configured by the AirPort Setup Assistant. [*]
After inserting the card, run the Setup Assistant. When it starts, you will see the dialog box in Figure
18-1.
[*]
Cards can also be configured with the System Preferences application. For completeness, this chapter discusses the Setup
Assistant first and the System Preferences application later in terms of monitoring and changing the configuration. However, there
is no reason why you cannot configure the card straight from the System Preferences application.
Figure 18-1. Initial AirPort Setup Assistant screen
Choose to configure the AirPort card and click Continue. The next step, shown in Figure 18-2, is to
select the network you wish to join. Every network within range is displayed in the pop-up menu.
Figure 18-2 shows the user selecting the secure.utah.edu network.
Figure 18-2. AirPort card network selection
Once the network is selected, you can move on to the third step: entering the nework password,
which is only useful for networks protected by WEP or WPA. 802.1X configurations must be set up in
Internet Connect, as described later in this chapter.
When connecting to a network, several different types of security can be selected. Figure 18-3 shows
the dialog box. The type of security is selected with the drop-down box from WPA Personal (WPAPSK), WPA Enterprise (WPA with RADIUS authentication), and WEP. To make matters easier for
users, Apple has allowed administrators to input WEP keys as variable-length passwords. The ASCII
text of the password is then hashed into a WEP key of the appropriate length. WEP keys can also be
entered in hexadecimal by prefacing them with a dollar sign, such as $EB102393BF. Hex keys are
either 10 hex digits (40-bit) or 26 hex digits (104-bit). Interpretation of the input string can be
forced to ASCII by enclosing the key in double quotes. [*]
[*]
For more information, see article 106250 in Apple's Knowledge Base at http://kbase.info.apple.com.
Configuring and Monitoring an AirPort Interface
You may need to change the AirPort configuration from time to time. You may move between
different 802.11 networks (ESSs) or need to change the WEP keys or IP settings. Once the card is
installed, you can change its configuration with the configuration tools provided with OS X. Apple's
configuration programs don't allow users to change any of the more complex 802.11 parameters. All
the information needed to join a network, for example, is broadcast in the Beacon frames. Apple
decided in most cases, it's enough to present the user with network names and prompt for security
settings.
Figure 18-3. AirPort network password entry
Basic configuration with the AirPort status icon
Once configuration is complete, the AirPort status icon is displayed in the upper-right corner of the
screen, next to the speaker volume, battery, and clock icons, provided you haven't turned off those
icons. The AirPort icon also indicates radio strength. In Figure 18-4, there are several solid
wavefronts on the icon. As you move farther from the access point and the signal degrades, the
number of bars decreases. When it is clicked, a drop-down command list offers the option of turning
the power to the AirPort card on or off, selecting or creating networks, and opening the Internet
Connect application to monitor the radio interface. It is quite handy for users to be able to turn off
the card at will. When you are out of range of a network, or just not using it, the card can easily be
powered down to save battery power.
Figure 18-4. AirPort status icon
In Figure 18-4, there are two networks within range: "Little Green Men" and "Luminiferous Ether."
The checkmark by "Little Green Men" indicates that it is the network to which the user is currently
connected, although the user can switch between the two networks simply by selecting a preference.
Any other network can be selected by using the "Other..." option and entering the name of the
network.
An IBSS can be created by going to the Create Network option and selecting the basic radio
parameters shown in Figure 18-5. The computer is set to create an IBSS with the network name of
Very Independent BSS; the radio channel defaults to 11 but can be changed to any of the 11
channels acceptable in North America and Europe. Every computer taking part in the IBSS must use
the same channel. IBSS networks can only be configured for WEP.
Figure 18-5. IBSS parameter setup
After you've set up an independent network, the system adds a new section titled "Computer to
Computer Networks" to the drop-down list, as shown in Figure 18-6. The AirPort status icon also
changes to a computer in the pie-wedge shape to indicate that the network is an IBSS rather than an
infrastructure network.
Figure 18-6. AirPort status icon while associated to an IBSS
Configuration with the System Preferences application
If you move between different ESSs, you can create a "location" for each and use this to configure
the ESS/password pair, which you can then pick from a menu as you move to a different location.
You can also preconfigure an ESS/password pair if you're not currently on the network for which you
are setting up.
The System Preferences application allows you to configure many system attributes, including those
that are network-related. Figure 18-7 shows the network panel of the System Preferences
application. The Show pop-up list can be set to any of the network interfaces in the system.
Naturally, when set to AirPort, it enables the fourth tab, as shown in the figure. The default tab is
TCP/IP (Figure 18-7), which can be set to configure interfaces manually or with the assistance of
DHCP or BootP. When set to DHCP, as in the figure, the leased address is shown. Although the DHCP
server on my network provided DNS servers, the server IP addresses are not shown. (They are,
however, placed in /etc/resolv.conf, as with any other common Unix system.)
The other network panel worthy of note is the AirPort tab (Figure 18-8), which can set the network to
join by default. In most cases, it should be left to Automatic, which enables the computer to search
through any networks in the general vicinity for one that has already been configured. Automatic
selection can reference security configurations built with Internet Connect, if necessary.
Monitoring the wireless interface
The wireless-interface status can be monitored with the Internet Connect application. The Internet
Connect application can be launched from either the Applications folder on the hard disk or from the
AirPort status icon. It can be used to display the signal strength of the nearest access point, as well
as change the network with which the station is associated. If the network is changed, it will
reference any 802.1X configuration necessary to authenticate to the network. See Figure 18-9.
802.1X on the AirPort
802.1X on Panther is set up through the Internet Connect application, which can be launched from
the drop-down menu under the AirPort status icon, or directly from the hard disk.
To see the 802.1X configuration, click on the lock labeled 802.1X. If the lock does not appear, either
select New 802.1X Configuration from the File menu, or press Command-Shift-X. Once 802.1X has
been displayed for the first time, it will always appear as an icon across the top (Figure 18-10).
To see the full configuration, select the Configuration drop down menu and choose Edit
Configuration..., which brings up the main configuration screen (Figure 18-11). The easiest task from
here is to edit the configuration. This brings up a configuration screen. Put in the username and
password, and select any authentication protocols that may be used. This screen allows selection of
the network interface to which the 802.1X credentials apply. In most cases, it will remain set to the
AirPort card, although wired 802.1X is gaining prominence. The username and password are
straightforward, as is the wireless network. By default, the drop down for Wireless Network will
display any encrypted networks detected by the AirPort, but it is possible to type in an arbitrary
SSID.
Figure 18-7. TCP/IP preferences tab of the Network Preferences settings
After typing in credentials, it may be necessary to configure the authentication method. The
supplicant has different configuration screens for each method, as discussed in subsequent sections.
After returning to the main screen of Internet Connect, you can press the Connect button to
associate to the network and attempt 802.1X authentication. The status bar at the bottom will go
through the following stages:
Figure 18-8. AirPort preferences tab of the Network Preferences settings
Idle
The AirPort is not associated to any network.
Connecting
The AirPort is associating with the selected SSID.
Authenticating
The AirPort is trading EAPOL frames with the AP or switch as it tries to validate user
credentials.
Connected via (EAP method)
If the authentication succeeds, the status bar will note that the system is connected, and it will
display the EAP method in use, as well as the time which the system has been authenticated.
Figure 18-9. Internet Connect monitoring application
Figure 18-10. 802.1X configuration in Internet Connect
Configuring EAP Methods
Each of the EAP methods has a different method configuration. Generally speaking, the Mac
supplicant hides a great deal of the complexity, unless it is required.
TTLS configuration
There are two potential TTLS configuration options, shown in Figure 18-12. The first, which is
mandatory, is the inner authentication type. In most cases, I expect this will be set to PAP, though it
is also possible to use CHAP, MS-CHAP, or MS-CHAP-V2. Like most supplicants, the Mac supplicant
supports hiding the user identity by configuring an anonymous outer identity.
Figure 18-11. 802.1X Configuration screen
Figure 18-12. TTLS configuration screen
PEAP configuration
On the Mac supplicant, only EAP-MSCHAP-V2 is supported as an inner authentication method. This
allows interoperability with user accounts stored as MD4 hashes or in cleartext. The main use for this
method is to support user accounts stored in a Windows network. On the configuration screen, Figure
18-13, the only option is to configure an outer identity to hide the user identity.
The Keychain
In Mac OS, passwords and certificates are stored on a keychain, which holds security-relevant
information in a centralized location. By logging in, the certificates and user credentials required for
network authentication are available to any application allowed by the user. Figure 18-14 shows a
simple keychain, which contains one RADIUS server certificate, the CA that signed that certificate,
and the 802.1X configuration for a network. The keychain provides access control to protect
configurations.
Figure 18-13. PEAP configuration screen
Figure 18-14. The keychain viewer
Adding to the keychain
When connecting to an 802.1X-protected network for the first time, any certificates that are not yet
trusted will cause the authentication to fail. Rather than failing the authentication, the supplicant will
present the certificates to the user, as shown in Figure 18-15. By inspecting the certificates, users
can make a decision about whether to trust them. More likely, a centralized information technology
department will test user credentials on the wireless network before turning the system over to the
end user, and can add any relevant certificates during the build process.
Figure 18-15. Adding a certificate to the keychain
Troubleshooting
When 802.1X authentication fails, the message that is received (Figure 18-16) does not provide a
great deal of information. An error code does not provide detailed diagnostic information about why
the failure has occurred.
Figure 18-16. X failure dialog box
Fortunately, the supplicant is capable of producing extensive debug logging. Create the directory
/var/log/eapolclient. If the directory exists and the NSDebugEnabled environment variable is set
when Internet Connect launches, it will log EAPOL frames from the user to
/var/log/eapolclient/uid(number)-(network interface).log. For example, if UID 501 attempts to use
802.1X on the AirPort (typically en1), the file will be uid501-en1.log. Internet Connect must be
launched from the command line to pick up the environment variable in the Terminal it is launched
from.
Mac:~$ sudo mkdir /var/log/eapolclient
Mac:~$ NSDebugEnabled=YES; export NSDebugEnabled
Mac:~$ /Applications/Internet\ Connect.app/Contents/MacOS/Internet\ Connect
The debug log will dump out every frame sent and received, in both ASCII and hexadecimal. It does
not print decrypted frames for EAP methods that use encryption. As an example, the following two
EAPOL-Key frames were received after a successful 802.1X authentication:
---------------------------------------2005/02/15 17:00:42.903841 Receive Packet Size: 75
Ether packet: dest 0:30:65:2:e7:36 source 0:b:e:a:70:2 type 0x888e
EAPOL: proto version 0x1 type Key (3) length 57
Signature: 6c 08 b3 97 b1 74 f2 ec 0e 2c 1f 66 ff 40 78 42 is valid
EAPOL Key Descriptor: type RC4 (1) length 13 Broadcast index 2
replay_counter: 42 11 48 e2 00 00 00 18
key_IV:
ca 7e d9 ff d4 47 80 ba 9e eb 33 c8 82 17 02 7c
key_signature: 6c 08 b3 97 b1 74 f2 ec 0e 2c 1f 66 ff 40 78 42
key:
fc 7a 7c 67 f6 f6 f5 7d a5 94 cb 49 1a
---------------------------------------2005/02/15 17:00:42.960436 Receive Packet Size: 64
Ether packet: dest 0:30:65:2:e7:36 source 0:b:e:a:70:2 type 0x888e
EAPOL: proto version 0x1 type Key (3) length 44
Signature: ce b3 45 05 47 72 5a 98 7c 64 0c d8 52 0d 8f 78 is valid
EAPOL Key Descriptor: type RC4 (1) length 13 Unicast index 0
replay_counter: 42 11 48 e2 00 00 00 19
key_IV:
45 e5 2a ad 1c 9c ea 8f 3c 58 a4 c4 6a e0 fa 82
key_signature: ce b3 45 05 47 72 5a 98 7c 64 0c d8 52 0d 8f 78
EAPOL: 2 bytes follow body:
0000 00 00
Mixing WPA and WEP on the Mac
As this book went to press, the Mac supplicant had a bug related to the interoperability
in environments which mix encryption types. It is relatively common to run dynamic WEP
and TKIP simultaneously. Stations that are capable of TKIP run it for unicast data, while
broadcast data is encrypted with dynamic WEP. Stations that are only capable of
dynamic WEP use it for both unicast and broadcast data.
In a mixed WEP/TKIP environment, Macintoshes are unable to connected to the network.
During the association process, the supplicant learns the network supports both WEP and
TKIP, and accepts the security parameters by associating. However, during the key
handshake, the supplicant insists on running only TKIP. With two different sets of
security parameters in the association process and the keying process, the standards
dictate that the AP should fail keying and kick the system off the network.
As a workaround, either reduce the security of the network to run only dynamic WEP,
run two parallel wireless networks (one WEP and one TKIP) so that Macs can use a TKIPonly network, or force all dynamic WEP devices off the network until they can be
upgraded to TKIP.
Chapter 19. Using 802.11 on Linux
When the first edition of this book was written, 802.11 was only just coming to Linux. Cards had to
be selected carefully because very few cards were supported with full open source drivers that
evolved at the same pace as the Linux kernel. Linux support has now moved into the mainstream,
with many vendors actively sponsoring driver development projects, or at the very least supporting
efforts to target their hardware. Broadcom is a notable exception.
Most 802.11 devices are supported by the PCMCIA system. As with Windows drivers, installing
wireless cards on Linux creates Ethernet interfaces. Many Linux drivers expose an Ethernet interface
through the kernel, and most drivers even name the resulting interfaces with the eth prefix.
Programs can use the Ethernet interface to send and receive data at the link layer, and the driver
handles Ethernet-to-802.11 conversions.[*] Many of the things you would expect to see with an
Ethernet interface remain the same. ARP works identically, and the IP configuration is done with the
same utilities provided by the operating-system distribution. ifconfig can even be used to monitor the
interface status and see the data sent and received.
[*]
There are two major encapsulation formats for data on 802.11. RFC 1042 is used for IP, and universally supported. Windows
and MacOS support IPX and AppleTalk with 802.1H. Not all Linux drivers support 802.1H.
PCMCIA Support on Linux
Initially, most wireless networking adapters were add-on cards that were based on the PC Card
specification.[ ] PC Cards attach to the system bus through a 16-bit controller interface that operates
at 8 MHz. Even though performance was limited by the bus to be roughly comparable to the 16-bit
ISA bus, it was more than capable of supporting relatively slow 802.11b wireless LANs. Higherperformance 802.11a and 802.11g cards require the performance of the next-generation CardBus
interface, which drives a 32-bit bus at 33 MHz. CardBus provides the dramatically improved
performance required by higher-bandwidth network interfaces. CardBus cards look nearly identical to
PC Cards, and are used in the same slots. Both CardBus and PC Cards are, not surprisingly,
configured and managed through the Linux PCMCIA utilities.
[
] The Personal Computer Memory Card International Association (PCMCIA) is the industry group that came up with the PC Card
specification. Initially, the cards were known as "PCMCIA cards," but the name was later shortened to "PC Card." I use both
interchangeably, especially since Linux support for these cards still goes by the longer PCMCIA. (In reference to the unwieldy
nature of the acronym, one common joke expands it as "People who Can't Manage Computer Industry Acronyms.")
Using external interfaces with antennas that protrude from the case of the laptop is not always
desirable. Laptop vendors have adopted a different form factor for "built-in" wireless devices called
Mini-PCI. Mini-PCI is a small version of the PCI interface that is only a few inches long. There no limit
on what could be made into a Mini-PCI card, but by far the most popular use of the slot is to add
wireless networking capabilties to laptops. Most Mini-PCI wireless cards consist of a PCI-to-Cardbus
bridge combined with a CardBus wireless LAN interface, and are therefore configured and managed
through the Linux PCMCIA system.
PCMCIA Card Services Overview
Card Services grew out of an attempt to simplify system configuration. Rather than dedicating
system resources to individual devices, the host system maintained a pool of resources for PC Cards
and allocated resources as necessary. Figure 19-1 shows the procedure by which cards are
configured on Linux.
When a card is inserted, the cardmgr process orchestrates the configuration of the device, as shown
in Figure 19-1 . The orchestration pulls together system resources, kernel components, and kernel
driver modules through the configuration files stored in /etc/pcmcia . Roughly speaking, the host
takes the following steps:
1. A card is inserted into an empty PC Card socket, and cardmgr is notified of this event. In
addition to any hardware operations (such as supplying power to the socket), cardmgr queries
the card information structure (CIS) to determine the type of card inserted and the resources it
needs. For more information on the CIS, see the sidebar "Card Information Structure " later in
this chapter.
2. cardmgr attempts to identify the card and load the appropriate kernel modules for support. The
PCMCIA card database stored in /etc/pcmcia/config is used to map the CIS database to drivers.
Drivers can also be loaded automatically through a hardware identification list that is part of the
module and stored in a module map (/lib/modules/(kernel version)/module.*map ). The main
task of the identification is to associate cards with a class . For the purposes of configuring
network cards, the important point to note is that items in the network class have additional
network configuration operations performed on them later. The card is identified by the CIS
data from step 1, and the class setting is set in the main system configuration file. At this point,
cardmgr beeps once. Successful identification results in a high-pitched beep; unsuccessful
identifications are indicated by a beep of lower pitch.
Figure 19-1. Linux PCMCIA configuration system
3. cardmgr determines which resources are available to allocate to the card. Blocks of system
resources are reserved for PCMCIA card use in the main configuration file, and cardmgr doles
out resources as needed to cards. The number of I/O ports needed and the size of the required
memory window are obtained from the CIS.
4. Resources allocated by cardmgr are programmed into the PCMCIA controller, which is depicted
in Figure 19-1 as interaction with the PCMCIA device driver. The Intel i82365SL PCMCIA
controller is the most common chip on the market, which is why the kernel module is shown as
"i82365." The new kernel-based PCMCIA system uses the yenta_socket driver instead of a
chipset-specific driver. PCMCIA controllers implement resource steering to map resources
required by the card onto available system resources. A card may ask for an interrupt, but the
actual assigned interrupt is irrelevant. In operation, the card simply asks the PCMCIA controller
to raise an interrupt, and the controller is responsible for looking up the interrupt assigned to
the socket and firing the correct interrupt line.
5. Part of the configuration information obtained from the lookup in step 2 is the name of the
device driver that should be loaded to use the newly inserted card. Drivers for PCMCIA cards
are implemented as kernel modules. As part of the insertion process, the driver is informed of
6.
5.
resources allocated in step 4. With proper module dependencies, module stacking can be used
to load multiple modules.
6. Further user-space configuration is performed based on the class of the device. Network cards,
for example, have further configuration done by the /etc/pcmcia/network script, which is
configured by editing /etc/pcmcia/network.opts . Successful configuration in this step generates
a second high beep, and failure is reported with a low beep. Additional configuration may also
be performed by the Linux hotplug system.
Interface names in Linux
When a driver creates the network interface, it is given a name consisting of a prefix and a number.
Many drivers will create additional network interface with a name of the form ethX , where X is the
next number in sequence. Most laptops now have built-in Ethernet interfaces that come up on boot
as eth0 , which makes many wireless interfaces start as eth1 . Older versions of the WaveLAN driver
created interfaces beginning with the prefix wvlan , but current versions use eth . Atheros-based
cards come up using the prefix ath .
Hotplug system for automatic configuration
To perform automatic configuration, Linux depends on the work of the hotplug configuration project,
whose web site is found at http://linux-hotplug.sourceforge.net/. When a device is inserted, the
hotplug system will call any needed PCI configuration commands before configuring the interface.
Hotplug scripts can help to automatically configure systems with large numbers of network
interfaces, or perform arbitrary configuration tasks. Many wireless devices implementing new
standards are under continued development after the product is released. New firmware updates to
cards can make new protocol features possible, or fix bugs in the released version. Hotplug can also
be configured to automatically update firmware when particular cards are plugged in.
Card Information Structure
When using dynamic WEP on Linux, the 802.1X supplicant derives WEP keys from the authentication
exchange, and then loads them into the driver. Up until 1999, many cards did not support the
dynamic provisioning of keys, and needed to reset the microcontroller in the card to make the new
key take effect. Drivers were written with this assumption, and the code that updated keys would
reset the hardware.
To enable automatic configuration, every PC Card has a blob of data that enables the card to describe
itself to the host system. The blob is called the card information structure (CIS) and takes a relatively
straightforward link-list format. The building blocks of the CIS are called tuples because they have
three components: a type code to identify the type of tuple, a length field, and a series of data bytes.
Tuple formats range from trivial to highly complex, which is why this book does not attempt to
classify them any further. Brave, stout-hearted readers can order the specification from the PCMCIA
and use the dump_cis tool on Linux to read the CIS of inserted cards.
The CIS assists the host in automatic configuration by reporting information about itself to the host
operating system. For example, network interface cards identify themselves as such, and the CIS
enables the Card Services software to allocate the appropriate resources such as I/O ports and
interrupt request (IRQ) lines. On Linux, the system administrator uses configuration files to match
the CIS data to the driver.
PCMCIA Card Services Installation
As of kernel 2.4, PCMCIA support has been integrated into the kernel, and was probably installed
with your distribution. It is not necessary to upgrade the kernel PCMCIA support. It may be
necessary to fetch drivers for your interface, but it is almost certain that the PCMCIA support is upto-date enough to handle any wireless interface you throw at it. (Older distributions can update
PCMCIA software through a distribution maintainer's package, or by building from the source at
http://pcmcia-cs.sourceforge.net/. )
Most configuration information is automatically detected from the environment without difficulty.
Depending on the distribution and its installation, it may be necessary to inform the PCMCIA software
which controller is running the PCMCIA slots by setting the PCIC variable in the system configuration
file. Kernel PCMCIA software sets the value to yenta_socket . Systems that use the standalone
PCMCIA support will set the variable to either tcic for Databook TCIC-2 chipsets, or i82365 for the
Intel i82365SL chipset. Slackware features a "probe" option to load modules until one successfully
loads.
Monitoring the Cards
The main tool used to control the PCMCIA subsystem is the cardctl command. It has several
subcommands that can be used to view information or configuration status. Each argument can also
take a slot number. Many laptop computers have two PCMCIA slots, in which case the slots are
numbered 0 and 1. An increasing number of notebooks designed for portability have only a single
slot, and therefore only have slot 0.
In addition to the hardware-based eject system, cardctl can be used to remove drivers by sending
the "card is removed" notification. Once the software has removed all the associated drivers and
powered the slot down, the card can be restarted by sending a software notification of the card
insertion. Having software control over the card allows the "reboot" of a card without having to
physically remove and reinsert it.
[root@bloodhound]# cardctl eject 0
[root@bloodhound]# cardctl insert 0
Drivers are selected based on the CIS data. To see the contents of the card's CIS, insert it and wait
for it to power up, then use the info or ident subcommands to cardctl to see the identification
information. Here is the information from a Proxim 8480 Gold card. Like nearly all cards with 802.11a
capabilities, it is based on the Atheros chipset. The card even reports itself as an Atheros reference
design:
root@bloodhound:~# cardctl info 0
PRODID_1="Atheros Communications, Inc."
PRODID_2="AR5001-0000-0000"
PRODID_3="Wireless LAN Reference Card"
PRODID_4="00"
MANFID=0271,0012
FUNCID=6
root@bloodhound:~# cardctl ident 0
product info: "Atheros Communications, Inc.", "AR5001-0000-0000", "Wireless LAN Referenc
manfid: 0x0271, 0x0012
function: 6 (network)
In addition to extracting the CIS identification with cardctl , the complete structure can be printed out
with cis_dump . In addition to identification information, the CIS includes information about the
supported speeds. In the case of the Proxim 8480, it reports exclusively on the 802.11a speeds, even
though it is capable of 802.11b-compatible operation:
root@bloodhound:~# dump_cis
Socket 0:
manfid 0x0271, 0x0012
config_cb base 0x0000 last_index 0x01
cftable_entry_cb 0x01 [default]
[master] [parity] [serr] [fast back]
Vcc Vnom 3300mV Istatic 25mA Iavg 450mA Ipeak 500mA
irq mask 0xffff [level]
mem_base 1
BAR 1 size 64kb [mem]
vers_1 7.1, "Atheros Communications, Inc.", "AR5001-0000-0000",
"Wireless LAN Reference Card", "00"
funcid network_adapter [post]
lan_speed 6 mb/sec
lan_speed 9 mb/sec
lan_speed 12 mb/sec
lan_speed 18 mb/sec
lan_speed 24 mb/sec
lan_speed 36 mb/sec
lan_speed 48 mb/sec
lan_speed 54 mb/sec
lan_speed 72 mb/sec
lan_media 5.4_GHz
lan_node_id 20 00 4d a6 d4 0a
lan_connector Closed connector standard
Socket 1:
no CIS present
Once the card has been inserted into the system, use the status subcommand to check on it. The
config subcommand will also report the system interrupt assigned by the system. In my case, the
laptop uses a PCI-to-Cardbus bridge, so the card inherits the IRQ assigned to the PCI bus.
root@bloodhound:~# cardctl status 0
3.3V CardBus card
function 0: [ready]
root@bloodhound:~# cardctl config 0
Vcc 3.3V Vpp1 3.3V Vpp2 3.3V
interface type is "cardbus"
irq 11 [exclusive] [level]
function 0:
The lights are not useful
Many drivers are intended to work with several different pieces of hardware, and each manufacturer
may customize the number and purpose of the lights on the card. Usually, there is one light to
indicate that the card is powered up, and a second light that will blink to indicate traffic from the
card. It may or may not blink in response to traffic from other stations. A few cards may have a third
light to indicate association status.
Linux drivers generally do not make an attempt to control the lights on each card in a vendor-specific
way, and internal cards may not have any lights at all. If your card behaves differently, don't panic.
Check at the access point for an association; if an association exists, then the card is functioning
normally but uses the lights in a different manner.
Troubleshooting Resource Conflicts
One of the revolutionary developments hyped by PCMCIA card vendors was that users were no
longer directly responsible for maintaining low-level hardware configurations on IBM-compatible
hardware. In many respects, this hype was overblown because users are still responsible for
maintaining the resource pools used by PCMCIA Card Services to draw from for automatic
configuration, and therefore they must still be familiar with the hardware configuration. Three major
resources are managed by Card Services for users: IRQ lines, I/O ports, and direct memory access
(DMA) channels. DMA channels are not required by network cards, however, and are not discussed in
this section.
IRQs
IRQs are used by devices that must use the CPU periodically. Interfaces use IRQs so that when a
buffer fills, the system CPU can be notified and drain the buffer. One limitation of the PC architecture
is that it has only 15 available IRQs, and many are occupied by standard hardware. Table 19-1 shows
common IRQ usage, which may help you determine which IRQs are available for PCMCIA cards.
Disabling any extra components frees the IRQ. Table 19-1 also shows common IRQ settings on PC
hardware. As a rule of thumb, IRQs 3, 5, and 10 are readily available on most machines.
0
System timer
Fires 18 times per second to maintain coarse clocking.
1
Keyboard
Allows operating system to monitor keyboard strokes by user.
2
Cascade
Two interrupt controller chips are used; the second controls IRQs 8-15 and is wired into IRQ 2 on the
primary.
3
Second serial port
The second and fourth serial ports (COM2 and COM4 under Windows) both use IRQ 3. If only one
serial port is present, IRQ 3 may be used by expansion devices.
4
First/third serial port
The first and third serial ports (COM1 and COM3 under Windows) both use IRQ 4. Generally, it is not
a good idea to use IRQ 4 for expansion devices because loss of the serial port also means that
terminal-emulation software cannot be used.
5
Second parallel port
Most systems have only one parallel port, but IRQ 5 is also commonly used for sound cards.
6
Floppy controller
All systems have floppy disks, which can be especially important on portable computers.
7
First parallel port
The first parallel port can frequently be disabled on laptops without an issue, unless the parallel port
is used extensively for printing.
8
RTC
The Real-Time Clock maintains finer-grained timers
9
Video (older systems)
Older systems required an IRQ for the video controller, and it was typically assigned to IRQ 9. Most
video controllers are now on the PCI bus and do not require a dedicated IRQ.
10
Usually available for expansion devices.
11
Usually PCI bus or SCSI controller
Generally not available for expansion devices.
12
Usually PS/2 mouse port
Generally not available.
13
FPU
The floating-point unit IRQ is used by the math coprocessor, even on systems with a CPU with an
integrated math coprocessor such as the Pentium series.
14
Primary IDE
The first IDE channel is used by the main system hard disk, and thus IRQ 14 is almost never
available on a portable system.
15
Secondary IDE
Portable systems typically place the CD-ROM on IRQ 15, making this IRQ unavailable for use.
Table 19-1. Common IRQ settings
IRQ number
Common usage
Purpose
I/O ports
I/O addresses are used for bidirectional communication between the system and a peripheral device.
They tend to be somewhat poorly organized, and many devices have overlapping defaults. Each I/O
port can be used to transfer a byte between the peripheral device and the CPU. Most devices require
the ability to transfer multiple bytes at a time, so a block of ports is assigned to the device. The
lowest port number is also called the base I/O address. A second parameter describes the size of the
I/O window. Table 19-2 lists some of the common port assignments. Refer to your hardware vendor's
documentation for details on additional devices such as IR ports, USB controllers, the PCMCIA
controller, and any resources that may be required by motherboard components.
Communication ports
First parallel port
0x3bc-0x3bf (4)
First serial port
0x3f8-0x3ff (8)
Second serial port
0x2f8-0x2ff (8)
Disk drives
Primary IDE
master: 0x1f00x1f7 (8)
slave: 0x3f60x3f7 (2)
Secondary IDE
master: 0x1700x177 (8)
slave: 0x376 (1)
Floppy controller
0x3f00x3f5 (6)
Input devices
Keyboard
0x060 (1)
0x064 (1)
Multimedia/gaming
Sound card
0x2200x22f (16)
FM Synth: 0x3880x38b (4)
MIDI: 0x3300x331 (2)
Joystick/Game port
0x2000x207 (8)
System devices
Interrupt controllers
0x0200x021 (2)
0x0a00x0a1 (2)
DMA controllers
DMA channels 0-3: 0x0000x00f (16)
Page registers: 0x0800x08f (16)
DMA channels 4-7: 0x0c00x0df
CMOS/real time clock
0x0700x073 (4)
Speaker
0x061
Math coprocessor
0x0f00x0ff (16)
Table 19-2. Common I/O ports
Device name
I/O range (size)
Linux Wireless Extensions and Tools
802.11 interfaces act like Ethernet interfaces once they are active, but they have many more
configuration options than Ethernet interfaces because of the underlying radio technology. Rather
than have each driver responsible for implementing its own configuration utility and reporting
mechanisms, the Wireless Extension API was developed to allow all drivers to implement common
functionality and use common configuration commands. Wireless extensions are also a great help to
application developers needing to get details from the interface because they can do it in a deviceindependent way. For example, xsupplicant makes use of wireless extensions to set WEP keys on an
interface without having to understand each driver's operations for setting keys. (At press time,
system calls for WPA keying were not yet part of the official wireless extensions release.)
Compiling and Installing
Wireless LAN extensions are enabled with the CONFIG_NET_RADIO kernel configuration option.
CONFIG_NET_RADIO-enabled kernels collect wireless statistics and expose additional data structures
used by most drivers. Most distributions that include kernels with wireless extensions also include the
wireless tools. If more recent versions are required, you may download them from
http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux, although it is probably easier to wait for
your distribution's kernel source to integrate the latest version and rebuild the kernel.
Interface Configuration with Wireless Tools and iwconfig
The main command-line tool for managing a wireless-extension-enabled driver is iwconfig, which is
designed to be a "wireless ifconfig." It can be used to set the radio interface parameters before using
ifconfig to set traditional network parameters.
When run with no parameters, iwconfig displays a list of system interfaces. Interfaces that are not
wireless will report that they have no wireless-specific data, while interfaces that do report wireless
data will dump out the data that they have. Not all drivers implement all features. When iwconfig is
run on a system with an old Orinoco card and an Atheros CardBus card, for example, the output will
probably resemble this:
[root@bloodhound]# iwconfig
lo
no wireless extensions.
eth0
no wireless extensions.
eth1
IEEE 802.11-DS ESSID:"11b-is-slow" Nickname:"HERMES I"
Mode:Managed Frequency:2.457GHz Access Point:00:E0:03:04:18:1C
Bit Rate:2Mb/s
Tx-Power=15 dBm
Sensitivity:1/3
RTS thr:off
Fragment thr:off
Encryption key:off
Power Management:off
Link Quality:46/92 Signal level:-51 dBm Noise level:-94 dBm
Rx invalid nwid:0 invalid crypt:0 invalid misc:0
ath0
IEEE 802.11 ESSID:"11a-is-very-fast"
Mode:Managed Frequency:5.28GHz Access Point: 00:0B:0E:00:F0:43
Bit Rate:36Mb/s
Tx-Power:off
Sensitivity=0/3
Retry:off
RTS thr:off
Fragment thr:off
Encryption key:E452-94AC-09DB-2200-1256-6D7D-74 Security mode:open
Power Management:off
Link Quality:31/94 Signal level:-64 dBm Noise level:-95 dBm
Rx invalid nwid:0 Rx invalid crypt:0 Rx invalid frag:0
Tx excessive retries:0 Invalid misc:0
Missed beacon:0
Both cards report the frequency of operation and some statistics about the quality of reception. The
"nickname" reported by the Orinoco card is not reported by the Atheros card because it is a feature
that is not implemented by the Atheros driver. (By default, the nickname is set to "HERMES I" after
the name of the Lucent chipset used in the Orinoco card.)
Wireless Extensions reports information on the link quality in two ways. It reports a noise floor and
signal level, and both are printed out by the driver. The signal to noise ratio is the difference between
them. In the previous example, the SNR on eth1 was 43 dB, and the SNR on ath0 is 31 dB. The Link
Quality statistic printed out by the driver reports the RSSI and the noise floor. In the previous
example, the card is reporting an RSSI of 31 dB over a noise floor of -94 dBm.
Finding networks
Some drivers support the use of wireless extensions to build a list of networks in the area. Rather
than configuration, use the iwlist command to get information from the card. As arguments, it takes
the interface and a subcommand. scan will print out the list of information discovered in the area.
Root privileges are required to retrieve scan data.
root@bloodhound:~# iwlist eth1 scan
eth1
Scan completed :
Cell 01 - Address: 00:07:50:D5:CE:88
ESSID:"LuminiferousEther"
Mode:Master
Frequency:2.417GHz
Quality:0/10 Signal level:-70 dBm
Encryption key:on
Bit Rate:1Mb/s
Bit Rate:2Mb/s
Bit Rate:5.5Mb/s
Bit Rate:11Mb/s
Cell 02 - Address: 00:09:5B:72:12:58
ESSID:"Mom's House"
Mode:Master
Frequency:2.467GHz
Quality:0/10 Signal level:-22 dBm
Encryption key:on
Bit Rate:1Mb/s
Bit Rate:2Mb/s
Bit Rate:5.5Mb/s
Bit Rate:11Mb/s
Cell 03 - Address: 00:0D:72:9B:FE:69
Noise level:-256 dBm
Noise level:-256 dBm
ESSID:"2WIRE086"
Mode:Master
Frequency:2.442GHz
Quality:0/10 Signal level:-28 dBm
Encryption key:on
Bit Rate:1Mb/s
Bit Rate:2Mb/s
Bit Rate:5.5Mb/s
Bit Rate:11Mb/s
Bit Rate:22Mb/s
Bit Rate:6Mb/s
Bit Rate:9Mb/s
Bit Rate:12Mb/s
Noise level:-256 dBm
Setting the network name
To connect to a network, the first task is to select the network that will be joined. That is done by
configuring the SSID with the essid[*] parameter. If the network name includes a space, it must be
enclosed in quotation marks. The network interface probably will not become active and start
searching for the network until you use ifconfig to activate it. Progress of the scan can be checked by
repeatedly running iwconfig and looking at the frequency. For example, the scan searching out an AP
on channel 56 might look something like this:
[*]
I use the term SSID in this book to refer to a network name. Some drivers, including the WaveLAN drivers, use ESSID instead.
The distinction is that an ESSID is a network name assigned to an extended service set, not any old service set.
[root@bloodhound]# iwconfig ath0
[root@bloodhound]# ifconfig ath0
[root@bloodhound]# iwconfig ath0
. . .
Mode:Managed Frequency:2.412GHz
. . .
[root@bloodhound]# iwconfig ath0
. . .
Mode:Managed Frequency:2.447Hz
. . .
[root@bloodhound]# iwconfig ath0
. . .
Mode:Managed Frequency:5.17GHz
. . .
[root@bloodhound]# iwconfig ath0
. . .
Mode:Managed Frequency:5.28GHz
. . .
essid "Space Cadet"
up
Access Point: 00:00:00:00:00:00
Access Point: 00:00:00:00:00:00
Access Point: 00:00:00:00:00:00
Access Point: 00:0B:0E:00:F0:43
When the card discovers the network, it will associate with the AP and report the AP's MAC address in
the iwconfig output. Associating with the network is merely the first step. Other configuration may be
necessary before the logical network connection.
Some drivers have problems if the channel is automatically configured. If the channel is hard set, the
driver may not search. To get a driver stuck on a channel to search, disable the card with cardctl
eject, reset the SSID, and then reenable the interface with ifconfig.
Setting the network channel
Different cards support different operating frequencies. On a modern Atheros-based card, the entire
ISM band is supported up to channel 14, and all three of the 5 GHz bands are supported as well. With
drivers that support the operation, it is possible to get a list of supported channels by using the iwlist
command:
root@bloodhound:~# iwlist ath0 channel
ath0
255 channels in total; available frequencies :
Channel 01 : 2.412 GHz
Channel 02 : 2.417 GHz
Channel 03 : 2.422 GHz
Channel 04 : 2.427 GHz
Channel 05 : 2.432 GHz
Channel 06 : 2.437 GHz
Channel 07 : 2.442 GHz
Channel 08 : 2.447 GHz
Channel 09 : 2.452 GHz
Channel 10 : 2.457 GHz
Channel 11 : 2.462 GHz
Channel 12 : 2.467 GHz
Channel 13 : 2.472 GHz
Channel 14 : 2.484 GHz
Channel 34 : 5.17 GHz
Channel 36 : 5.18 GHz
Channel 38 : 5.19 GHz
Channel 40 : 5.2 GHz
Channel 42 : 5.21 GHz
Channel 44 : 5.22 GHz
Channel 46 : 5.23 GHz
Channel 48 : 5.24 GHz
Channel 50 : 5.25 GHz
Channel 52 : 5.26 GHz
Channel 56 : 5.28 GHz
Channel 58 : 5.29 GHz
Channel 60 : 5.3 GHz
Channel 64 : 5.32 GHz
Channel 100 : 5.5 GHz
Channel 104 : 5.52 GHz
Channel 108 : 5.54 GHz
Channel 112 : 5.56 GHz
Current Frequency:5.28GHz (channel 56)
The operating frequency can be selected in three ways. Most cards will hunt for an SSID if it is given
to them as described previously. For cards that do not support scanning, the freq parameter can take
an operating frequency directly, or the channel parameter can be used with the appropriate channel
number, and the driver will derive the frequency from the channel number. The following two
commands are equivalent:
[root@bloodhound]# iwconfig ath0 freq 2.432G
[root@bloodhound]# iwconfig ath0 channel 4
Setting the network mode and associating with an access point
Most 802.11 stations are in either ad hoc networks or infrastructure networks. The iwconfig
nomenclature for these two modes is Ad-hoc and Managed. Select between them by using the mode
parameter:
[root@bloodhound]# iwconfig ath0 mode Ad-hoc
[root@bloodhound]# iwconfig ath0 mode Managed
For stations in an infrastructure network, the ap parameter may be used to request an association
with the specified MAC address. However, the station is not required to remain associated with the
specified access point and may choose to roam to a different access point if the signal strength drops
too much:
[root@bloodhound]# iwconfig ath0 ap 01:02:03:04:05:06
Setting the data rate
Most cards support multiple bit rates. iwconfig allows the administrator to choose between them by
using the rate parameter. Bit rates can be specified after the rate parameter, or the keyword auto
can be used to specify that the card should fall back to lower bit rates on poor-quality channels. If
auto is combined with a bit rate, the driver may use any rate lower than the specified rate:
[root@bloodhound]# iwconfig ath0 rate auto
Configuring static WEP keys
The key parameter controls the WEP function of the driver. Keys can be entered as hexadecimal
strings using the key subcommand. Enter the string without any delimiters, four digits at a time with
dashes between the two-byte groups, or in the colon-separated byte form. The following commands
are equivalent:
[root@bloodhound]# iwconfig ath0 key 0123456789
[root@bloodhound]# iwconfig ath0 key 0123-4567-89
[root@bloodhound]# iwconfig ath0 key 01:23:45:67:89
Many drivers support the use of 104-bit WEP keys as well. 104 bits is 13 bytes, or 26 hexadecimal
characters:
[root@bloodhound]# iwconfig ath0 key 12345678901234567890123456
[root@bloodhound]# iwconfig ath0 key 1234-5678-9012-3456-7890-1234-56
[root@bloodhound]# iwconfig ath0 key 12:34:56:78:90:12:34:56:78:90:12:34:56
Although multiple keys can be entered using a bracketed index number, this capability is not used on
many networks. Dynamic WEP key assignment through 802.1X is a much easier way to ensure the
use of multiple keys, and is much more secure.
[root@bloodhound]# iwconfig ath0 key 0123-4567-89
[root@bloodhound]# iwconfig ath0 key 9876-5432-01 [2]
[root@bloodhound]# iwconfig ath0 key 5432-1678-90 [3]
Once multiple keys have been entered, select one by entering the index number without a key value:
[root@bloodhound]# iwconfig ath0 key [2]
Activate WEP processing using key on and disable WEP using key off. These can be combined with an
index number to select a new WEP key:
[root@bloodhound]# iwconfig ath0 key [3] on
[root@bloodhound]# iwconfig ath0 key off
Finally, two types of WEP processing can be done. An open system accepts data frames sent in the
clear, and a restricted system discards cleartext data frames. Both of these parameters can be
combined with an index number:
[root@bloodhound]# iwconfig ath0 key [4] open
[root@bloodhound]# iwconfig ath0 key [3] restricted
The key parameter may also be accessed with the encryption parameter, which may be abbreviated
to as few characters as enc. I prefer to use key because it seems clearer to me, but you may choose
whichever you like.
Tuning 802.11 parameters
iwconfig allows you to tune the RTS and fragmentation thresholds. The RTS threshold of most drivers
is 2,347, which effectively disables RTS clearing. In an environment likely to have hidden nodes, it
can be set using the rts_threshold parameter with iwconfig. rts_threshold can be abbreviated as rts.
[root@bloodhound]# iwconfig wvlan0 rts 500
The default value of the fragmentation threshold is 2,346. In noisy environments, it may be worth
lowering the fragmentation threshold to reduce the amount of data, which must be retransmitted
when frames are lost to corruption on the wireless medium. Set the parameter by using the
fragmentation_threshold argument to iwconfig. It may be set anywhere from 256 to 2,356, but it
may take on only even values. fragmentation_threshold may be abbreviated as frag.
[root@bloodhound]# iwconfig ath0 frag 500
802.11 stations maintain several retry counters. When frames are retransmitted "too many" times or
wait for transmission for "too long," they are discarded. Two retry counters are maintained. The long
retry counter, set by the retry parameter, is the number of times transmission is attempted for a
frame longer than the RTS threshold. The short retry counter, set by the retry min parameter, is the
number of times transmission will be attempted for a frame shorter than the RTS threshold. Unlike
many drivers, iwconfig also allows for configuration of the maximum frame lifetime with the retry
lifetime parameter. To specify a value in milliseconds or microseconds, append "m" or "u" to the
value:
[root@bloodhound]# iwconfig ath0 retry 4
[root@bloodhound]# iwconfig ath0 retry min 7
[root@bloodhound]# iwconfig ath0 retry lifetime 400m
Agere (Lucent) Orinoco
Wireless networking is far older than the 802.11 standard. Several proprietary approaches had been
released and were gaining a foothold in the market in the early 1990s. One of the first notable
products was the NCR WaveLAN, released when NCR was a division at AT&T.[*] When Lucent was
spun off from AT&T in 1996, the WaveLAN division, like most communications product manufacturing
at AT&T, was made a part of Lucent.
[*]
NCR, which was founded in 1884 as the National Cash Register Company, was acquired by AT&T in 1991. I can only assume
that a cash register company was interested in wireless networking because it would enable cash registers to be placed anywhere
without wiring.
Early WaveLAN hardware was a completely proprietary system. After 802.11 was finally standardized
in 1997, new hardware that complied with the standard was sold under the WaveLAN brand. To
distinguish the standards-compliant cards from the proprietary cards, the former were called
"WaveLAN IEEE" cards, while the latter were simply "WaveLAN" cards.
As the market for 802.11 hardware continued to develop, Lucent decided to rename the WaveLAN
division under another brand name. The new name, Orinoco, comes from the third largest river
system in the world. During the rainy season in South America, the Orinoco swells with fresh rainfall
flowing in from over 200 tributaries. At its peak, the river grows to over 10 miles wide and more than
300 feet deep. Nearly 1,000 miles of the Orinoco's 1,300 miles are navigable; it is no wonder that
the river's name is derived from the native words for "a place to paddle."
Lucent's initial strategy for support on open source platforms was to offer a choice of drivers. A
closed-source proprietary binary driver, wavelan2_cs , provided full functionality, and a second, less
functional open source driver, wvlan_cs , was made available under the GPL. wvlan_cs was based on
a low-library provided with the closed source driver, and it was an evolutionary dead-end. Rather
than continue along that path, wvlan_cs was rewritten with new low-level operations and became
orinoco_cs in Linux kernel version 2.4. (orinoco_cs was originally known as dldwd_cs , which stood
for David's Less Dodgy WaveLAN Driver!)
In addition to the WaveLAN cards and any OEM versions of WaveLAN cards, orinoco_cs contains basic
support for some PRISM-2-based cards and Symbol cards that use the same MAC chipset. orinoco_cs
has been part of the Linux kernel distribution since kernel Version 2.4.3.
Compiling and Installing
All distributions based on the 2.4 and later kernels include orinoco_cs . In many cases, it will be
possible to use the driver that shipped with your kernel. The major exception is if you plan to run
xsupplicant and need to apply the key management patch, which will be discussed later.
PCMCIA configuration
Older distributions may ship with the wvlan_cs driver still enabled. To change the driver used by the
distribution, it is sufficient to change the module binding the PCMCIA configuration. The author of
orinoco_cs supplies a file, hermes.conf , which contains card definitions for the cards supported by
orinoco_cs . Because hermes.conf ends in .conf , it is sourced by the line at the end of
/etc/pcmcia/config that reads all .conf files. However, to avoid binding conflicts, you must comment
out all the lines that bind the older wvlan_cs driver to newly inserted cards. Alternatively, it is
sufficient to edit the definition of your wireless card to bind the orinoco_cs driver after grabbing
identification information from the output of dump_cis :
# in hermes.conf
#
card "Lucent Technologies Wavelan/IEEE"
version "Lucent Technologies", "WaveLAN/IEEE"
bind "orinoco_cs"
# from standard /etc/pcmcia/config
#
# card "Lucent Technologies WaveLAN/IEEE"
#
version "Lucent Technologies", "WaveLAN/IEEE"
#
bind "wvlan_cs"
Doing it yourself
This is part of the kernel, so you generally don't need to install it unless you're looking for a
particular bug fix or feature enhancement. The most common reason to recompile the driver is a
requirement to support dynamic WEP with 802.1X authentication. xsupplicant distributes a patch to
be applied for rekeying. Firmware updates may also be necessary to support dynamic keys. I have
used firmware version 8.42. Later versions almost certainly work, but earlier versions may not.
Requirements are straightforward. The code itself is fetched from the distribution site at
http://ozlabs.org/people/dgibson/dldwd/. Make sure that you have a copy of patch to apply the
patches of interest (the rekey patch is not required in version 0.15rc2):
gast@bloodhound:~$ cd orinoco-0.13e
gast@bloodhound:~/orinoco-0.13e$ patch -p1 <
../rekey_patch_orinoco-0.13e
patching file orinoco.c
patching file orinoco.h
msg@bloodhound:~/orinoco-0.13e$ make
(build messages snipped)
root@bloodhound:/home/msg/orinoco-0.13e# make install
if [ -d /etc/pcmcia ]; then install -m 644 -o 0 -g 0 hermes.conf /etc/pcmcia/hermes.conf;
mkdir -p /lib/modules/2.4.26/kernel/drivers/net/wireless
for f in hermes.o orinoco.o orinoco_cs.o orinoco_plx.o orinoco_tmd.o orinoco_pci.o; do \
if test -e /lib/modules/2.4.26/pcmcia/$f; then \
install -m 644 -o 0 -g 0 $f /lib/modules/2.4.26/pcmcia/$f; \
else \
install -m 644 -o 0 -g 0 $f /lib/modules/2.4.26/kernel/drivers/net/wireless/$f; \
fi; \
done
depmod -a
At the end of the build process, the system has several kernel modules: orinoco_cs.o , the PCMCIA
interface; orinoco.o , the hardware driver; and hermes.o , the driver for the MAC chip.
Configuring the orinoco_cs Interface
Configuration of orinoco_cs is identical to the configuration of wvlan_cs . When the card is inserted,
the /etc/pcmcia/wireless script is run, using the configuration options in /etc/pcmcia/wireless.opts .
The wireless script is a frontend to the iwconfig program. Editing fields in wireless.opts sets the
arguments to iwconfig . For details on configuring the options to iwconfig , see the previous section
on the wvlan_cs driver.
Resetting on WEP Key Installation
Not all card manufacturers have committed to supporting Linux. Revealing intellectual property by
giving away driver source code is usually the major concern. Finding the engineering time to write,
maintain, and support drivers given to the user community is another.
When using dynamic WEP on Linux, the 802.1X supplicant derives WEP keys from the authentication
exchange, and then loads them into the driver. Up until 1999, many cards did not support the
dynamic provisioning of keys, and needed to reset the microcontroller in the card to make the new
key take effect. Drivers were written with this assumption, and the code that updated keys would
reset the hardware.
Resetting the hardware when a key is installed is undesirable behavior with dynamic WEP. Once the
authentication has completed, a key must be installed. If the driver resets after key installation, the
link will bounce. In many cases, the AP will note the link bounce and require a new authentication.
Once the supplicant detects the link state change, it will run the authentication again to derive a new
key. After the new authentication completes, the supplicant resets the card to install the new key,
starting the cycle over yet again.
Most interface cards have firmware upgrades available which eliminate the need to reset the
hardware after key installation. Once the right version of firmware is in place, the driver needs to be
updated to a dynamic WEP-capable version that does not send reset commands.
If dynamic WEP is not working because the card resets after the authentication succeeds, check
either the source code or the online forums for the driver for your card to see if an update or patch is
required.
Atheros-Based cards and MADwifi
802.11a is the choice for high-speed, high-density 802.11 networks, and most 802.11a devices on
the market use Atheros chipsets.[*] The driver project is called the "Multiband Atheros Driver for
WiFi," or MADwifi ; the project's home page is http://sourceforge.net/projects/madwifi/.
[*]
It is likely that any card with 802.11a support is based on an Atheros chipset. Rather than include a list that will quickly become
outdated, I suggest you refer to the searchable list at http://customerproducts.atheros.com/customerproducts/ .
Although best known for 802.11a support, there are several Atheros chips available. There are three
generations of chipset. The initial chipset, the 5210, only supported 802.11a. It was followed by the
first dual-band chipset, the 5211, which added 802.11b support. Most devices now on the market use
the 5212, which is a dual-band/tri-mode chipset that can support 802.11a, 802.11b, and 802.11g. In
addition to the different radio support, later generations are capable of performing much more
complex security operations. All chipsets are supported by MADwifi.
Driver Architecture and the Hardware Access Layer (HAL)
MADwifi is split into two parts: an open source driver, and a closed-source library called the
Hardware Access Layer (HAL). Atheros chipsets are extremely flexible, and capable of tuning outside
the unlicensed bands because they use a limited form of a software-defined radio.
Atheros has interpreted FCC regulations as requiring that the HAL be kept closed. (Broadcom
interprets the rules in the same way, and they have a similar chipset on the market.) FCC rules
require that software-defined radios must not be modifiable by the user to operate outside of their
certified frequency bands. Devices that do not enforce the terms of their certification are potentially
subject to sanctions. If devices sold for use in the unlicensed spectrum were suddenly able to disrupt
licensed communications, vendors of such devices might also be penalized.[*]
[*]
Christian Sandvig, a professor at the University of Illinois at Urbana-Champaign, takes exception to the SDR interpretation of
FCC rules in a paper titled "Hidden Interfaces to 'Ownerless' Networks"
(http://www.spcomm.uiuc.edu/users/csandvig/research/Hidden_Interfaces.pdf ).
For commercial products, ensuring that the software stays in the unlicensed spectrum is easyjust
avoid compiling and distributing code that breaks the rules. Users cannot edit compiled code, so they
are unable to break the rules. Open source software changes the game. Distributing code that follows
the rules does not prevent modification that break the rules.
Atheros faced a choice: ship a driver with completely open source code and risk losing regulatory
approval to sell chips, or find some way of protecting the radio spectrum. Obviously, a driver exists
because they took the latter course. Rather than have the open source driver interface directly with
the radio chipset, it accesses the radio through the HAL's API. The HAL is generic enough that it is
used unchanged by many other operating systems developing open source drivers; there are zero
dependencies on the host system software, though it does depend on the host hardware's instruction
set. The HAL is available for x86 architectures (both 32- and 64-bit), ARM, MIPS, PowerPC, and
XScale. Although closed-source code is often viewed with suspicion in the open source community, I
consider it a sign of corporate commitment to the Linux platform that Atheros was willing to support
the development of the driver.
Requirements
MADwifi makes extensive use of wireless extensions, and is consistently developed against the latest
version. Either use a recent distribution with current wireless extensions support, or upgrade the
kernel to a recent distribution.
Drivers often make use of other kernel functionality, and MADwifi is no different. Building MADwifi
depends on having kernel headers and configuration files for the running kernel. If you have built a
custom kernel for your hardware, by definition you have kernel source and configuration. Not all
distributions will install the kernel source and configuration by default; you may have to find the right
package.
UUCP tools will also be necessary to get the HAL. To safely transfer the binary HAL over arbitrary
network paths, it is uuencoded. MADwifi's make script calls uudecode. Depending on your
distribution, uudecode may be part of the shell archive utilities package (sharutils ), or it may be part
of a UUCP tools package.
Building the Driver
MADwifi is still under active development, and it changes regularly. Rather than having periodic
packaged releases, it is constantly maintained in CVS. Fetch the latest version from CVS:
root@bloodhound:~# cvs -z3 -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/madwifi co ma
When the command completes, the source code and (encoded HAL files) will be downloaded into a
directory madwifi . To build the driver, use the standard open source command set:
root@bloodhound:~# cd madwifi
root@bloodhound:~/bloodhound# make
(lots of build messages)
root@bloodhound:~/bloodhound# make install
The main modules are wlan.o , ath_pci.o , and ath_hal.o . Several other cryptographic modules are
loaded on demand to support particular encryption schemes. If the MADwifi build system cannot
detect the right place in the filesystem to install to, you may need to manually place the required
modules in the correct location and run depmod -a to update modules.
Using the Driver
Current versions of the MADwifi driver publish a list of supported cards to the kernel, which is
revealed when module dependencies are built. The MADwifi driver supported device list includes
several devices with the manufacturer ID of 0x168c, which belongs to Atheros. Provided that module
dependencies are built correctly, there is no need to modify the PCMCIA configuration scripts.
Loading the driver creates an interface prefaced with ath . Most likely, it will be ath0 , since you will
only have one Atheros card running in your system. Older versions of MADwifi used the prefix wlan .
If you see wlan0 , update the driver.
MADwifi is correctly integrated with the Linux hotplug system. When the module is inserted, the
hotplug system will attempt to register the interface. It is almost certain that the start-up scripts in
place on your distribution will need to be edited. Open wireless networks can probably have light
editing; using 802.1X requires heavier editing because the interface must be authenticated before
starting DHCP.
Using Windows Drivers Under Linux
In addition to xsupplicant, there is one commercial program available from Meetinghouse Data
Communications. Meetinghouse's AEGIS client supports a wide variety of EAP methods (MD5, TLS,
TTLS, PEAP, and LEAP). Due to the difficulties in keeping up with new distributions, however, it is
only officially supported on RedHat versions 8 and 9. The Linux version of the Meetinghouse
supplicant does not support WPA for the same reason xsupplicant does not: the lack of a standard
WPA keying API.
Not all card manufacturers have committed to supporting Linux. Revealing intellectual property by
giving away driver source code is usually the major concern. Finding the engineering time to write,
maintain, and support drivers given to the user community is another.
Drivers written natively for Linux are preferable for a variety of reasons. However, if your card
vendor does not yet have a Linux driver, you can use the Windows driver through a wrapper
program. Two software packages can run the Windows NDIS driver in a sandbox and translate
Windows system calls into their Linux equivalents. The NDISWrapper project
(http://ndiswrapper.sourceforge.net/ ) is a completely open source implementation. A company
called Linuxant provides a commercial program called DriverLoader
(http://www.linuxant.com/driverloader/ ) for the same purpose. Trial licenses are free, but ongoing
use does require payment.
802.1X on Linux with xsupplicant
I have been fortunate enough to volunteer at the Interop Labs wireless security initiatives for the
past few years, where I get to experience interoperability close up. In 2004, developers from the
Open1X project participated in the interoperability test event, bringing open source code into an
event that had previously been the preserve of product manufacturers. As the popularity of wireless
networks has waxed, so has interest in solutions for connecting Linux systems to secure wireless
LANs. xsupplicant is one of the leading implementations of 802.1X for Linux.
Requirements
Before even considering working with xsupplicant, the most important task is to ensure that 802.11
is already working! Many "wireless" problems are really PC Card problems. Obviously, you will not be
able to attach to an authenticated network until building and configuring xsupplicant, but you should
be able to insert your wireless card and have it recognized by the system. Card Services will
generally sound a high-pitched beep when a card is inserted to indicate it is recognized, resources
were successfully allocated, and the driver was loaded. (A second low-pitched beep may follow
indicating a configuration failure, but that is okay. Chances are that you want your ultimate
configuration to run xsupplicant, so it's acceptable at the outset to not have configuration support
immediately.) With the driver successfully loaded, you should be able to run iwconfig and obtain
statistics from the driver.
Even if the driver for your chosen card is working, make sure it has support for dynamic WEP.
Although there is some dispute over the absolute security level of dynamically keyed WEP systems, it
is unquestionable that dynamic WEP has a significantly higher level of security than a single-key WEP
system. It is a virtual certainty that any card you want to use with xsupplicant should support the
use of dynamic keys. Most modern cards, those released in early 2003 and later, have driver support
for dynamic WEP already, but a few older cards require patches. The popular Hermes-based Orinoco
802.11b cards fall into the latter category. Depending on the driver, you may also need to rebuild
your kernel (or at least get a copy of your kernel's configuration) in order to compile a driver.
Finally, xsupplicant needs one major library. As discussed previously, most EAP methods are based
on TLS in one way or another. xsupplicant uses the OpenSSL TLS implementation. Install version
0.9.7 or later. Before fetching the source from openssl.org , check to see if there is a package
available for your distribution.
Compiling and Installing xsupplicant
Compiling and installing xsupplicant follows the same routine that any other open-source package.
Released code can be fetched from http://sourceforge.net/projects/open1x/ , and compiled in the
standard Unix way:
root@bloodhound:~# tar -xzvf xsupplicant-1.0.1.tar.gz
root@bloodhound:~# cd xsupplicant
root@bloodhound:~/xsupplicant# ./configure
root@bloodhound:~/xsupplicant# make
root@bloodhound:~/xsupplicant# make install
As this book went to press, xsupplicant 1.0.1 was several months old, and massive development had
occurred in the CVS version. (This section is written using a late 2004 CVS version.) Among many
other improvements, the CVS version has preliminary WPA support. The CVS version can be fetched
from an anonymous CVS server, and compiled in the standard Unix way:
root@bloodhound:~# cvs -z3 -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/xsupplicant c
root@bloodhound:~# cd xsupplicant
root@bloodhound:~/xsupplicant# ./configure
root@bloodhound:~/xsupplicant# make
root@bloodhound:~/xsupplicant# make install
As a result of the build, you get three executables installed; the only one you are likely to use is
/usr/local/sbin/xsupplicant .
Configuring xsupplicant
When run, xsupplicant searches for its configuration file in /etc . The config file, /etc/xsupplicant.conf
, does not get installed by default, so copy it over to the expected location.
root@bloodhound:~/xsupplicant# cp etc/xsupplicant.conf /etc/
Configuration files specify the authentication method, user identity, possibly the password, and a
certificate location to validate the certificate from the network. Certificate authentication of the
network may be disabled by setting the certificate location to NONE, but it is not recommended.
When setting up the network, you may need to convert certificates between different forms.
OpenSSL can easily convert between different formats by specifying the formats on the command
line.
root@bloodhound:~# openssl x509 -inform DER -outformPEM -in MyCA.der -out MyCa.pem
Passwords are stored in the configuration file in a nonencrypted form, so you may wish to have the
system prompt for the password to avoid putting sensitive information in configuration files. When no
password is in the file for the network you are connecting to, xsupplicant will prompt the user.
In the configuration file, settings can be stored as part of a profile for an SSID. The following
example is a simple configuration for a network using PEAP, with EAP-MSCHAP-V2 inner
authentication on a network called batnet . xsupplicant can be configured to run a command after
authentication completes, or that can be left to the configuration scripts on the operating system.
### GLOBAL SECTION
logfile = /var/log/xsupplicant.log
network_list = all
default_netname = batnet
first_auth_command = dhcpcd %i
###
NETWORK SECTION
batnet{
# allow_types = eap_tls, eap_md5, eap_gtc, eap-otp
allow_types = eap_peap
# Phase 1 ("outer") identity
identity = msg
# Alternative, but common specification is to not reveal username
# identity = anonymous
eap-peap {
# It is a good idea to validate the certificate and not do "none"
# root_cert = NONE
root_cert = rootCA.pem
root_dir = /etc/xsupplicant.d
chunk_size = 1398
random_file = /dev/random
# ** Inner method configuration
allow_types = eap_mschapv2
# Inner method configuration
eap-mschapv2 {
username = msg
password = imnottelling
}
}
}
Pseudorandom number generation
Like many other security protocols, EAP methods require a "good" source of random numbers. Linux
provides two random number devices, /dev/random and /dev/urandom . The former returns random
bits that are within the system's entropy pool. If insufficient entropy is available, the read will block
until sufficient entropy is available, while the latter will return as much data is requested, perhaps at
the cost of quality.
Connecting and Authenticating to a Network
The first step in gaining access to an 802.1X-protected network is to associate to it. xsupplicant
depends on the system to perform the association. The initial configuration should also configure the
card to use encrypted frames, although the key configured does not matter. Configure a dummy key
to indicate to the driver that it should run in encrypted mode; xsupplicant will replace the key after
successful authentication with a wireless extension library call. The most common way of configuring
the connection is to use iwconfig to plumb a key and configure the network, and then use ifconfig to
bring the interface up and start searching for the network. Naturally, these commands may be
triggered by the configuration system scripts when a wireless card is inserted.
root@bloodhound:~# iwconfig ath0 key 12345678901234567890123456
root@bloodhound:~# iwconfig essid essid "batnet"
root@bloodhound:~# ifconfig ath0 up
After the system has associated with the network, run xsupplicant. The -i option indicates which
interface it should run on. Debugging is activated with -d , with the following letters determining
what information is printed out. Log messages are sent to the console with -f . The following example
has cut the raw packet dumps.
root@bloodhound:~# /usr/local/sbin/xsupplicant -i ath0 -dasic -f
Using default config!
network_list: all
Default network: "default"
Startup command: "echo "some command""
First_Auth command: "dhclient %i"
Reauth command: "echo "authenticated user %i""
Logfile: "/var/log/xsupplicant.log"
Allow Types: ALL
ID: "msg"
peap root_cert: "NONE"
peap chunk: 1398
peap rand: "/dev/random"
PEAP Allow Types: ALL
mschapv2 username: "msg"
mschapv2 password: "imnottelling"
Interface ath0 initalized!
At this point, xsupplicant has read the configuration file, and checks to see if the wireless interface is
associated to an AP. If it is, the authentication process can begin.
[INT] Interface ath0 is wireless!
[INT] The card reported that the destination MAC address is now 00 0B 0E 00 F0 40
[INT] Working with ESSID : batnet
[CONFIG] Working from config file /etc/xsupplicant.conf.
[STATE] (global) -> DISCONNECTED
[STATE] Processing DISCONNECTED state.
[STATE] DISCONNECTED -> CONNECTING
[STATE] CONNECTING -> ACQUIRED
[STATE] Processing ACQUIRED state.
Once the system has associated to an AP, it begins authenticating.
Connection established, authenticating...
[STATE] Sending EAPOL-Response-Identification
[STATE] ACQUIRED -> AUTHENTICATING)
[STATE] Processing AUTHENTICATING state.
[STATE] Sending EAPOL-Response-Authentication
****WARNING**** Turning off certificate verification is a *VERY* bad idea!
[AUTH TYPE] Packet in (1) :
20
[AUTH TYPE] Setting Key Constant for PEAP v0!
[INT] Interface eth0 is NOT wireless!
Userdata is NULL! We will probably have problems!
[STATE] (global) -> DISCONNECTED
[STATE] Processing DISCONNECTED state.
[STATE] DISCONNECTED -> CONNECTING
[STATE] Processing AUTHENTICATING state.
[STATE] Sending EAPOL-Response-Authentication
You should not
At this point, a great deal of debugging output will appear as certificates are exchanged back and
forth and validated by both sides of the conversation. Once the TLS tunnel is established, the debug
output shows the inner authentication. In the case of EAP-MSCHAP-V2, the RADIUS server sends a
challenge. Based on the challenge value and the shared secret, a response is generated.
[AUTH TYPE] (EAP-MSCHAPv2) Challenge
[AUTH TYPE] (EAP-MS-CHAPv2) ID : 2F
[AUTH TYPE] Authenticator Challenge : C6 02 26 BE C3 E0 44 03 13 6E 1F BA F0 B3 1D 5A
[AUTH TYPE] Generated PeerChallenge : 28 62 AA A2 8C 8E EB 82 D1 9B 2F 9A 54 67 93 2C
[AUTH TYPE] PeerChallenge : 28 62 AA A2 8C 8E EB 82
[AUTH TYPE] AuthenticatorChallenge : C6 02 26 BE C3 E0 44 03
[AUTH TYPE] Username : msg
[AUTH TYPE] Challenge : 48 E7 AA 53 54 52 98 62
[AUTH TYPE] PasswordHash : C5 A2 37 B7 E9 D8 E7 08 D8 43 6B 61 48 A2 5F A1
[AUTH TYPE] Response : 4D 96 51 8C 18 3A F7 C7 70 15 47 13 19 D8 D6 9B 36 00 AD E8 FA 9A 0
[AUTH TYPE] myvars->NtResponse = 4D 96 51 8C 18 3A F7 C7 70 15 47 13 19 D8 D6 9B 36 00 AD
[AUTH TYPE] response->NT_Response = 4D 96 51 8C 18 3A F7 C7 70 15 47 13 19 D8 D6 9B 36 00
[AUTH TYPE] (EAP-MSCHAPv2) Success!
[AUTH TYPE] Server authentication check success! Sending phase 2 success!
Once the system has successfully authenticated, the AP will supply keys. Keys are both encrypted
and authenticated, using keys derived from the shared cryptographic keys from the phase 1 TLS
exchange. Two key messages are sent, one with the unique unicast key for the station, and one with
the broadcast key shared by all stations. xsupplicant uses the wireless extensions API to set the keys
in the driver, where they will be viewable with iwconfig .
Processing EAPoL-Key!
[INT] Key Descriptor
= 1
[INT] Key Length
= 13
[INT] Replay Counter
= 41 2F BB 2D 00 00 00 D5
[INT] Key IV
= 69 4C 45 D7 CF C3 DD CD 2A 3A F3 CB 04 7A F4 A3
[INT] Key Index (RAW) = 01
[INT] Key Signature
= C2 05 6C 3A EB 25 E9 B9 8E FC 60 D6 77 44 57 22
[INT] EAPoL Key Processed: broadcast [2] 13 bytes.
[INT] Key before decryption : ED 5D 03 D2 7A DE B4 60 29 FD FD F5 42
[INT] Key after decryption : FB BB AC D3 6F 7D 0A 3F FF 2A CF 33 4E
[INT] Successfully set WEP key [2]
Processing EAPoL-Key!
[INT] Key Descriptor
= 1
[INT] Key Length
= 13
[INT] Replay Counter
= 41 2F BB 2D 00 00 00 D6
[INT] Key IV
= 66 15 69 E2 B2 8C 0E 89 7C D3 94 8C 93 25 43 1B
[INT] Key Index (RAW) = 80
[INT] Key Signature
= 49 C1 15 B8 E9 D0 87 53 A6 FD 5D 76 CB 51 9D 65
[INT] EAPoL Key Processed: unicast [1] 13 bytes.
[INT] Using peer key!
[INT] Successfully set WEP key [1]
[INT] Successfully set the WEP transmit key [1]
Some drivers implement private system calls to report keys. Atheros cards using the MADwifi driver
can list keys with iwlist ath0 key ; many drivers will also report the unicast key in iwconfig .
Commercial 802.1X on Linux
In addition to xsupplicant, there is one commercial program available from Meetinghouse Data
Communications. Meetinghouse's AEGIS client supports a wide variety of EAP methods (MD5, TLS,
TTLS, PEAP, and LEAP). Due to the difficulties in keeping up with new distributions, however, it is
only officially supported on RedHat versions 8 and 9. The Linux version of the Meetinghouse
supplicant does not support WPA for the same reason xsupplicant does not: the lack of a standard
WPA keying API.
WPA on Linux
WPA is not generally available on Linux because there is no standard way of obtaining the necessary
information to perform the four-way handshake. xsupplicant runs in user space, but some of the data
necessary to compute the four-way handshake must be obtained through the driver. Each driver has
defined its own system calls for fetching this data, which requires supplicants to track the
development of private system calls in every driver. A future release of the wireless extensions API
will support the necessary system calls, and WPA will be made available in both open source and
commercial products.
Chapter 20. Using 802.11 Access Points
In even the simplest 802.11 network, proper configuration of the access points is essential. Without
properly configured network interfaces, no traffic will be bridged on to the wired network.
In the early days of 802.11, access points were simple devices. They had a pair of interfaces, and
connected wireless devices to an existing wired LAN. Early APs were bridges between the wireless
and wired realms, with few additional features. When the first generation of wireless LAN products hit
the market, there was a broad separation into cheaper home gateways and expensive business-class
products. Products in both classes performed the same functions, although the latter typically used
higher-capacity components to provide flexibility and investment protection, as well as the first crude
large-scale management tools.
At the dawn of 802.11, APs were initially standalone independent devices. In the early days, wireless
LANs offered the ability to have untethered access, but lacked serious security and management
capabilities. Although these islands of networking occasionally communicated with each other, they
were not often used as part of a large-scale network system because the protocols to enable them to
do so had not been developed or tested on a large scale. As wireless networks became more popular
and grew larger, flaws in the traditional standalone access point model became apparent.
Sensing a market opportunity, several vendors have built a second generation of wireless network
hardware. Newer approaches concentrate management and support functions for APs into an chassis
that has the basic functions of an Ethernet switch, which enables the APs themselves to be dumbeddown radio interfaces. Although there are several approaches, the broad category of "lightweight" or
"thin" access points plus a control system is often labeled a wireless switch. A great deal of interest
has surrounded the switch-based architecture since it was first popularized in late 2002, but the
fundamental technology must perform the same functions as first-generation access points.
This chapter discusses, in general terms, how to use access points. It takes a look at a full-featured
standalone device, the Cisco 1200 series AP. Although there is a great deal of power in the Cisco
1200, it can be formidable to configure. For a look at the lower end of the market, this chapter looks
at the Apple AirPort Express.
General Functions of an Access Point
Broadly speaking, there are two price classes of access points in the marketplace. A low-cost tier
consisting of home devices is sold widely through retail channels directly to the end user. These lowcost devices are often specialized computing platforms with only limited memory and storage.[*] The
higher-cost tier incorporates additional features required to support large deployments; frequently,
these devices have additional memory and storage and incorporate more general-purpose hardware.
The difference between the two price tiers is that the higher-cost devices are meant to work together
as a system to build a much more reliable, secure, and manageable network. To use a somewhat
simplistic analogy, the small-scale wireless LANs that have proliferated in homes and small offices
are like cordless phones. They extend a single network out over a limited range, and that is all.
Large-scale wireless LANs are much more like cellular telephony, with a strong focus on maintaining
a network connection in a much more demanding environment. Frequent user motion and and handoff between APs is a given, as is a much higher standard of management and troubleshooting tools.
[*]
For cost savings, many of the low-cost run a stripped-down version of Linux, and have been the subject of a great deal of
software hacking. See, for example, the HyperWRT (http://www.hyperwrt.org) and wifi-box (http://sourceforge.net/projects/wifibox) projects.
Cutting across the different market segments, however, is a set of generic features that are required
to fulfill the service promises made in the 802.11 standard. Configuration of these features, of
course, is vendor-specific, but many products are fairly similar to each other in purpose and design.
Most obviously, access points are bridges between the wireless world and the wired world. As
bridges, then, all access points have features that one would expect to see on a network bridge. They
have at least two network interfaces: a wireless interface that understands the details of 802.11 and
a second interface to connect to wired networks. I am not aware of any access point that does not
use Ethernet as the wired back-end, though it is certainly not required by any part of the standard.
As wireless LANs have grown up, more of the high-end access points have begun to support VLANs
on the network uplink. Lower-end access points may have a "WAN" port, which is usually a second
Ethernet port for use with a cable modem or DSL, though I have seen a few products that have RS232 serial ports to support dial-up modems.
All wireless interfaces must provide basic support for the 802.11 channel access rules, but the
similarity ends there. Early access points implemented the entire 802.11 protocol at the edge of the
network; many newer devices have moved some of the 802.11 processing away from the edge of the
network and have split the 802.11 MAC across multiple system components. Most access points offer
the ability to use external antennas to fine-tune range and coverage area.
Bridges have some buffer memory to hold frames as they are transferred between the two interfaces,
and they store MAC address associations for each port in a set of internal tables. Bridging tables are,
of course, highly implementation-specific, and there is no guarantee of similarities across the
industry. The most basic and inexpensive devices will usually assume that they are the only access
point in the network, and bridge accordingly. When access points support roaming, it may be
necessary to move sessions and user data in between access points. High-end access points may
need to augment a basic bridging table with VLAN information on the wired interface, as well as
information about how users authenticate their connections.
Commercial-grade devices are also designed to work cooperatively; the most common feature is a
vendor-proprietary method to move association data from access point to access point without
interrupting link-layer connectivity. Network management is generally much more sophisticated on
commercial-grade products to enable network engineers to manage the tens or hundreds of devices
used to create a large-scale coverage area.
Initially, management through a TCP/IP network interface was a standard feature. One of the big
innovations in the past few years has been the development of "thin" access point solutions that
move management functions from access points to central concentration devices.
Depending on the market for which an access point is developed, it may offer services to its wireless
clients. The most popular service is DHCP; wireless stations may be assigned addresses automatically
upon association. Larger-scale devices often rely on existing DHCP servers on the network to ensure
consistence across access points. Many access points can also perform network address translation
(NAT), especially the "home gateway"-type products that can connect to a modem and dial up an
ISP.
Security has been a sore point for wireless network managers since before the advent of 802.11's
success. Access points have a privileged position with respect to security concerns because they are
the gateways to the wired network and are ideally positioned to implement security policies. In
addition to first-generation security approaches such as MAC address filtering, most products now
implement stronger user-based authentication. Wi-Fi Protected Access (WPA) can be run with a preshared key in most home products, and with an external authentication server in large corporate
deployments. Many high-end devices now offer significant integration with the existing wired
network. Using those features to best extend the wired network will be discussed in the next chapter.
Management interfaces often leave something to be desired. Configuration of access points tends to
be challenging because access points must be manufactured cheaply, and low-cost devices tend not
to have the processing power to run an easy-to-use configuration engine. Most vendors use
lightweight operating systems running on low-powered hardware, but one of the trade-offs of using a
lightweight operating system is that it does not provide the programming environment necessary to
build rich functionality. Early access points offered both a command-line interface and a web-based
management interface. The recent development of "Wi-Fi switches" offers some hope for network
administrators. Rather than requiring management of individual APs as standalone network
elements, stripped-down (or "thin") access points are managed through a handful of centralized
control switches. With greater processing power and functionality, the switches can support more
functionality and much improved management interfaces.
Debugging and troubleshooting tools are as advanced as management tools, which unfortunately
means that they often leave network administrators mired in inconclusive or irrelevant information.
Ideally, products should maintain detailed logs of activities, but it is common to find vague logs of
results that give very little insight into failures. Counters can be helpful, but only if the right counters
are accurately maintained. Tools such as ping and traceroute are common, but network analyzers
and packet capture tools are not.
Types of Access Points
Broadly speaking, there are three major types of access points. Many of the best-known devices are
low-cost access points sold at major consumer electronics outlets. Although these devices make up
the bulk of the market, they are unsuited for use in a large-scale deployment. Just as consumer
electronics-class Ethernet switches are not suitable for building a major network, cheap APs cannot
offer the features needed to build a major wireless netowrk. Higher-priced devices with significant
additional functionality exist for the corporate enterprise market.
For the home: residential gateways
The low-cost tier is composed of devices often called residential gateways. Residential gateways are
designed to be as low-cost as possible, so only the basic features required for the typical small or
home office are included. To further reduce cost, most of the residential products are based on
"reference designs" from 802.11 chipmakers. Equipment manufacturers may (or may not) customize
a reference design, the external case, and sell the resulting device under their own brands.
Residential gateways generally share the following characteristics:
Most devices include a DHCP server to make plug-and-play configuration easier.
They are often deployed by users with one routable IP address, so NAT implementations are
common.[*] Many can use PPPoE or DHCP to dynamically assign the routable external address.
[*]
The NAT implementation is usually restrictive. It is able to translate many internal devices to varying ports on the
external IP address, and fixed ports on the external IP address to specific internal addresses (for, say, inbound web or
SSH requests). Some vendors may refer to this as port address translation (PAT) instead of NAT.
Depending on the type of customer the residential gateway is aimed at, the WAN interface is a
modem, a serial port, or even DSL. (Some residential gateway products may use an Ethernet
port as the "WAN" connection to a cable modem or DSL modem.)
They are often built as a single integrated unit, complete with a built-in antenna. If suitable
coverage cannot be found, it is necessary to relocate the entire unit.
Many products now claim to have an IPsec pass-through feature to allow the use of IPsec
through NAT, which works with varying degrees of success depending on the IPsec VPN solution
chosen.
Configuration of residential gateways usually relies on a default internal IP address. When it is
plugged in and powered up for the first time, you connect with a web browser to its default
address and enter the default username and password. In some cases, all of a manufacturer's
devices may come up with the same address by default. A popular choice for the default web
address is often 192.168.0.1, the first address in the RFC 1918 reserved address block of
traditional class C address. As wireless LANs have become more popular, however, some
vendors have built equipment to coexist with other gear by default, either by choosing a
random value for the final number in the IP address, or by testing to ensure the address is not
in use.
They are often sold directly to the end user and are designed to be aesthetically pleasing.
Unfortunately for many end users, the improved visual design sometimes prevents the stacking
of residential gateways with other network equipment. Many vendors design their equipment to
be stackable with their other components, however.
Security configuration options are often limited to smaller-scale solutions. Much older residential
gateways only implemented MAC filtering or static WEP, so be careful about purchasing used
equipment. Nearly every device now sold uses WPA's preshared key authentication with
dynamic encryption keys.
Most residential devices do not have sophisticated radios. They typically have just a single radio
interface, running either 802.11g or 802.11a. The former is more common in residential devices
because the 2.4 GHz frequency used by 802.11g has greater coverage. Residential networks are
typically limited by their uplink to the Internet, and do not need to be built with small coverage
areas for dense, high-bandwidth coverage.
As this book was written, residential gateways typically cost $35 to $100. Common manufacturers
are D-Link, Linksys, and Netgear. Apple's AirPort is sometimes placed in this category as well,
although it is priced significantly higher and has more features.
For the office: enterprise access points
Enterprise gateways, which often go by many other names that imply the buyer values features over
cost, provide everything residential gateways do, plus additional features useful for larger-scale
environments. Enterprise gateways generally share the following characteristics:
The area over which mobility is required is much larger and requires several access points
working in concert. Enterprise products support some sort of protocol to move sessions between
access points.
Enterprise products are built around upgradeability to offer a service life as long as possible.
They are typically built on relatively high-powered generic hardware, and implement a great
deal of higher-layer functionality in software that can be easily upgraded. They may have radios
that are easy to swap out. Early enterprise APs used PC Card radios, so a switch from 802.11b
to 802.11a involved changing a card, and possibly a software upgrade. As the price of
components has fallen, radio cards are typically no longer upgradeable, but the software is.
One of the advantages to the high level of software control of an enterprise-grade AP is that
new security developments can often be added in with a software update. Any serious
enterprise-grade AP will support WPA, and most vendors promised easy software upgrade paths
to 802.11i. As new security features are standardized, they can be added on to an existing
network without changing hardware. Enterprise-grade APs often add security in layers, so new
security mechanisms can be used in conjunction with older mechanisms. Hardware-based
enhancements to security often appear in more expensive devices first, too. When chipsets
started supporting AES acceleration in hardware, it appeared on high-end devices first. Some
enterprise APs can support multiple security standards simultanously.
Enterprise-type deployments are frequently intended to create a coverage blanket throughout a
relatively large area. Power is not always convenient to the location of an access point, but it is
typically abundant in the wiring closet. All enterprise-grade APs support drawing power over
Ethernet using IEEE 802.3af.
The radio side of a high-end AP is frequently much more sophisticated than a low-cost AP. Most
enterprise APs have an antenna connector that allows the attachment of a variety of external
antennas to tailor the coverage area and coverage quality to your needs. Some APs are
beginning to incorporate highly advanced antenna technology that allows very fine control over
transmission patterns, or allows a single antenna to be used by multiple MAC chips.
Transmission power can usually be adjusted to enlarge or shrink the coverage area as desired.
Many enterprise APs can also support multiple virtual radio networks over the antenna.
In addition to radio tunability, enterprise APs can support some form of virtual access points.
Multiple SSIDs can be configured, and each can be assigned its own authentication and
encryption settings. This feature is often used to create parallel networks with different security
settings to accommodate older equipment while offering maximum security for newer
equipment.
Enterprise APs are designed to integrate into an existing network. On the security side, that
means that they often must integrate into an existing security architecture, usually by plugging
in to an existing user database. Once users are successfully authenticated to the network,
though, they must take advantage of network services. A second integration point is the way
that an enterprise AP extends existing access control and user privileges that may already be
present on the wired network.
Integration with an existing network also extends to the data plane as well. Most enterprisegrade APs are designed to perform dynamic VLAN assignment based on user attributes from the
authentication server.
Frequently, site survey tools come bundled with enterprise-class products so network managers
can plan large deployments by directly assessing coverage quality. These tools vary greatly in
their sophistication. Some site survey tools are little more than a historical statistical readout,
while extremely advanced tools can derive a network layout from the physical environment.
Reflecting the administrative demands, configuration of enterprise-class devices is done with
easily scripted command-line interfaces or SNMP, and monitoring and management capabilities
are far more extensive than in residential gateways. Some products are also controlled through
large-scale management frameworks that enable a single administrator to monitor and change
configuration on hundreds or even thousands of access points.
Enterprise gateways are often deployed in packs. Aesthetic requirements in typical office space
may require unobtrusive mounting, so these devices offer flexibility in the way that they mount
in the building. Many are designed to be mounted above ceilings, and may be plenum rated.[*]
Devices installed in air ducts and air-handling spaces must meet strict smoke emission
requirements for safety reasons. Plenum-rated APs can be installed nearly everywhere; devices
that lack the necessary certifications may be restricted in possible mounting locations.
[*]
For information on flame tests, see http://www.houwire.com/catalog/technical/cable_flame.asp.
Naturally, these additional capabilities do not come without a price. Most enterprise-grade APs list for
$500 to 1,000, though they are often available at significant discounts. Over time, the price of highend APs does fall, but it is not subject to the same downward pressure as residential-class products.
The canonical example of an enterprise-grade AP is the Cisco 1200 or Cisco 1100. Both are built on
relatively generic hardware, run a full-blown version of the Internetwork Operating System, and are
given new features on a regular basis. Proxim, Symbol, 3Com, and HP produce competing products
with similar feature sets.
For the large office: wireless switches
One of the biggest changes in the time since the first edition of this book was published is the
emergence of the "wireless switch" or "thin AP" architecture, in which relatively lightweight access
points are controlled by a centralized switch. The driver behind the thin AP or wireless switch
architecture is increased efficiency over first-generation products. Part of the efficiency is based on
the technology itself. Thin AP architectures remove processing from the AP and move it to an
aggregation device. Eliminating processing at the AP removes components and cost from the AP,
increasing service lifetime. If configuration is removed from the AP as well, there are fewer managed
elements in the network.
Centralizing capabilities in the controller can also lead to increased flexibility. For the same cost,
concentrated hardware in the controller can provide more processing power than distributed
processing at the access point. Coordinating activity between access points allows network managers
to load-balance clients between APs, monitor radio activity centrally, and extend the existing network
more easily.
The cost of a wireless switch-based solution may depend a great deal on its size. Most vendors offer
a variety of controllers, which may range from just a few APs up to hundreds of APs. The original
switch solution was Symbol's Mobius product; solutions were later built from the ground up by
Airespace, Aruba, and Trapeze.
Power over Ethernet (PoE)
If your network will be of any size at all, power over Ethernet (PoE) is extremely important. When
designing a wireless network, access points should be located where it makes sense to put them for
radio propagation purposes, not where it makes sense to put them for electrical power purposes.
Access points sold in the first wave of products had to be powered like most other computing
devicesthat is, plugged into a nearby AC power outlet. To make matters worse, many of them used
large "wall wart" transformers that blocked adjacent outlets, too. Locating access points was not a
matter of finding the best locations for providing service, but near nearby power outlets. Modifying
access point locations can be challenging if one of the constraints is staying near electrical power.
Due to the dangers of electrical work, there are legal requirements to use licensed electricians in
many locations. Some organizations may have also negotiated with unions to provide skilled
electrical labor. In either case, modifying access point locations may be quite expensive. In many
locations, data cable such as Ethernet can be moved or installed by network technicians, so APs can
be installed or relocated by nonunion labor.[]
[]
Some jurisdictions may define Ethernet as "low-voltage wiring," which requires that it be installed by an electrician. The city of
Redmond, Washington, taxes low-voltage wiring as part of the building permit process.
Types of PoE
There are a number of types of power equipment on the market; many comply with the IEEE 802.3af
standard. 802.3af defines a transmission standard that can supply 48 volts DC up to a maximum of
approximately 15 watts at the full length-reach of an Ethernet cable. The standard itself can be
downloaded from the IEEE Get802 site, and more information is available from
http://www.poweroverethernet.com/.
The most basic distinction is which wires in the Ethernet cable supply the power. Data is always
carried on the data conductors (wire pairs 1 and 2 and 3 and 6). Alternative A in the standard carries
power on the data conductors, while Alternative B carries power on the unused pairs (wire pairs 4
and 5 and 7 and 8). When power is carried on the data cables, the voltage polarity may go either
way, as shown in Figure 20-1. The corresponding pinout is in Table 20-1. Power carried on the
unused data cables must have a specified polarity. Nearly all vendors use positive polarity on the first
pair; the notable exception is early Cisco-proprietary power devices, which reversed polarity from
most other devices.
Figure 20-1. Power wire map
Table 20-1. Wire usage in 802.3af
Wire number
Pair mapping
Alternative A (MDI-X)
Alternative A (MDI)
1
1 - data
-
+
2
1 - data
-
+
3
2 - data
+
-
4
3 (unused)
+
5
3 (unused)
+
6
2 - data
7
4 (unused)
-
8
4 (unused)
-
+
Alternative B
-
A further distinction between types of power devices is where the power is supplied. Power
equipment may be present in the switch, which is referred to as endpoint power. Alternatively, it may
be added into the wiring by a power injector, in which case it is called mid-span equipment. Most
802.11 installations to date have used power injectors, and therefore mid-span equipment. As switch
vendors have come out with power blades and they have been much more widely purchased,
endpoint injection is become much more common. Figure 20-2 illustrates the difference between the
two options. The injector shown in the mid-span case may be a stand-alone device, or it may be a
"power patch panel" with multiple ports. Gigabit data links can only be endpoint powered, although
this is not yet an important distinction for 802.11.
Figure 20-2. Endpoint and mid-span power injectors
802.3af incorporates detection of powered devices to avoid frying equipment plugged into a power
port. Early PoE equipment would supply the full voltage all the time, and had a propensity to destroy
devices that were not designed to cope with power. For safety reasons, 802.3af includes a handshake
that slowly ramps up power. If there is no response to the initial probe from the power equipment, it
stops providing it to protect the device at the end.
Power injectors almost always have two copper Ethernet ports and a power cord. One Ethernet port
is for the network connection, and the second provides power on the network connection on the first
port. A notable exception is the Cisco AIR-PWRINJ-FIB, which provides power on an Ethernet port
while using a 100BASE-FX fiber for network connectivity.
Selecting Access Points
When choosing an access point, you should take a number of factors into account. With the
emergence of 802.11 as the main vendor-neutral standard, standards compliance is generally not a
big factor. Any serious vendor hoping to build large networks must promise compatibility with all
existing standards, and will often need to commit to supporting important future landmark standards.
Several organizations have launched test programs to certify interoperability. The best-known
program is the Wi-Fi Alliance's Wi-Fi certification, though other organizations are seeking to capitalize
on the popularity of wireless LANs by launching competing programs. The most rigorous test
program in the industry is run by the University of New Hampshire's Inter-Operability Lab (IOL). The
IOL test program is not as widely known as the commercial programs because it is intended to help
build better products, not to provide marketing "seals of approval" to vendors. As always, when you
study certification programs, ensure that the vendor has gone through the relevant tests that are of
interest to you.
Security is often a prime concern when deploying a wireless network. With WEP thoroughly broken,
you almost certainly want to select an access point that has security capabilities based on 802.1X
and 802.11i. Small installations can use pre-shared key authentication, while more securityconscious installations should opt for full RADIUS integration. In larger environments, it may also be
worthwhile to select a product that can act as multiple virtual access points as well as mix encryption
and authentication types.
In some deployments, getting power to the access points can be a major headache. To blanket an
area with the coverage required for a large implementation, access points often need to be placed in
an area where power is not easily accessible. Long antenna runs can degrade signal quality
unacceptably, so it is much better to bring power to the location. Installing new electrical conduits is
often quite expensive. Work must be performed by licensed electricians, and building codes may
impose additional restrictions. Many products can supply power over the Ethernet cable. Cheaper
products often have proprietary power equipment, while more expensive products can use standardsbased power equipment. Network wire is not subject to the same restrictions as electrical cable and
can be installed by network administrators.
Some environments may require providing coverage over both indoor and outdoor areas. External
antennas are often useful for creating a dense coverage blanket over an area, or for beaming the
signal in a particular direction over a courtyard. Not all access points can connect to external
antennas; it may be an extra-cost option. Even if an access point has an external antenna connector,
there is no guarantee that you'll be able to find a wide range of antennas available for use. 802.11
only requires that any connectors for external antennas have a standard impedance of 50 ohms. If
external antennas are important for your deployment plans, make sure that a wide range of antennas
is available, whether through the 802.11 vendor or another source. Outdoor installations may require
environmentally "hardened" access points or enclosures as well.
Environmental considerations may also play a role in certifications required for APs. Equipment
installed in air-handling spaces must be plenum rated. Certified plenum-rated devices are lit on fire
and the resulting smoke is tested for clarity. Opaque smoke might be forced through a ventilation
system and distributed throughout a building, which would obscure emergency exit signs and would
be a hazard. Plenum-rated devices are made of materials that do not release dense smoke. The
smoke is often quite hazardous. Plenum-rated APs are safer because they will not obscure exits, not
because the smoke is nontoxic. Devices that are installed above the ceiling do not generally need to
be plenum rated unless the above-ceiling area is an air return space. Buildings with air-handling
ducts do not generally need plenum-rated APs, but many building inspectors will mistakenly require
them. Local building codes may also require them.
If roaming is important, you will need to sort through a variety of different technologies and
approaches, with the added wrinkle that roaming may not always perform well between different
vendors' products. The best way to ensure interoperability between vendors is to select a system that
bases roaming and handoff on dynamic VLAN assignment. As stations authenticate, they will retain
the same logical point of attachment to the network.
802.11 includes a number of powersaving functions in the standard. Most are optional. If your
deployment is based heavily on battery-powered devices, it may be worth evaluating which
powersaving features are included with particular devices. It may also be worth experimenting with
devices to see just how much longer batteries last with the powersaving functions enabled.
Device management is an important consideration. Wireless networks are a new service, and
network staff will need to plan, evaluate, purchase, deploy, and maintain the additional hardware.
Large deployments may have tens or hundreds of access points, which can easily make network
management a headache without good tools. Does the vendor offer an access point manager to
configure large numbers of devices in parallel? Can management of the access points be incorporated
into your existing network management infrastructure using tools that you already have deployed?
Are the management tools secure enough? Many products can be managed only with cleartext
protocols, which may be just an annoyance or a major violation of a security policy. Experience with
other network devices has shown that software upgrades are a frequent occurrence. How is the
software upgraded, and how much functionality can upgrades add? Can new protocol features be
added with firmware updates?
Depending on the size of the deployment, it may be possible to evaluate equipment before buying.
See if you can get a feel for the range of each access point and test with a variety of common cards.
Capacity on an 802.11 network is ultimately limited by the radio link, but you will want to make sure
that there are no other capacity restrictions. Does the access point provide the processing power to
run the wireless side at maximum capacity with security protocols enabled? Not all products
incorporate cryptographic acceleration for all protocols. Products that depend on a central
cryptographic processor to run security systems may run out of capacity if they are upgraded to
faster PHY standards, and may also suffer if the uplink is insufficient. Fast Ethernet suffices for
802.11a/g-based networks today, but it will likely limit the performance of radios built on the future
802.11n standard. Try to set up a test network and get a feel for the configuration required to
integrate the access points with the rest of your network gear.
As with many other purchasing decisions, of course, there are a number of "soft" factors that may
not be easily quantifiable. Warranties, a relationship with the vendor, and the quality of the technical
support may all influence the purchasing decision. Soft factors are not technical nor easily
quantifiable, however, so I will not discuss them.
Are Access Points Really Necessary?
Access points are not required for a wireless network. Wireless stations can be used in independent
networks, which do not require an access point. Building a Unix box that routes between an Ethernet
network and a wireless network is not difficult, and hardware can often be reused from the scrap pile.
Why, then, would anybody use an access point?
Now that most access points have fallen well below the $100 mark, building a Unix router is no
longer a cost-effective option for single-access point networks. Once you consider what your time is
worth, building a Unix router is a pretty silly use of time. Access point hardware has some
advantages over redeployed general-purpose platforms, too. Access points are small devices with no
moving parts. As a result, they do not consume a great deal of electrical power and do not generate
much heat. There is one notable exception to this rule, though. Apple offers a "software base station"
that transforms any desktop machine into a bridging access point. With a few mouse clicks and very
little effort, a desktop computer can become a base station.
Unix-based routers have never been effective in larger deployments because of the lack of mobility
support. Effective roaming requires transparent bridged access, not routed access, to the link layer
at different physical locations. However, roaming with 802.11 is possible only when access points can
communicate with each other to track the movement of a wireless station. In the future, it is likely
that an open source Unix distribution will have the features necessary for an access point: low-level
access to fundamental 802.11 parameters on the card, Ethernet bridging, and an IAPP. Until then,
though, there is no substitute for commercial products.
Unix-Based Access Points
One of the most basic preconditions for making a Unix-based access point is enabling
access point functions in the wireless interface card. One of the major hurdles is
rewriting the 802.11 headers. All traffic in an infrastructure network flows through the
access point. Access points must rewrite the transmitter and receiver addresses in the
802.11 headers. Other management functions may be required as well. For example,
802.11 includes a number of powersaving mechanisms for infrastructure networks in the
specification, but they can be used only on networks with access points that implement
them.
There is also a nontechnical hurdle. Many vendors have actively supported the
development of open source Unix drivers for their cards. After all, vendors make money
selling hardware, and it is a good thing for them to sell cards for all client systems, even
those that run open source Unix. Access points are a different story, however. Vendors
have not been as forthcoming with the interface used to put cards into the access point
mode. Access points are quite lucrative, and providing a driver interface in the access
point mode in the card could potentially cannibalize access point sales.
At one point, the only way to get an Intersil-based card to act as an access point
interface was to purchase the reference design from Intersil. (The reference design
shipped with firmware that had access point functionality, and that firmware was not
sold separately.) Intersil's shipping station firmware does, however, include something
called a "Host AP Mode." In the Host AP Mode, the PRISM chipset automatically takes
care of "menial" tasks, such as transmitting Beacon frames and acknowledging incoming
transmissions. Jouni Malinen has developed a driver to use the Host AP Mode with Linux.
In conjunction with the Ethernet bridging implementation in the kernel, this driver can be
used to build an access point with full 802.1X and WPA support. It is available from
http://www.epitest.fi/Prism2/.
With the present state of driver software, it is possible to build a Unix-based router. (I
mean "router" pedantically, as "layer 3 network device.") One interface would connect to
a wired network as it always has, and a second wireless interface could be run in IBSS
mode. Ross Finlayson has established a community network at a coffee house in
Mountain View, California using a FreeBSD-based router. The project's home page is at
http://www.live.com/danastreet/, and there is a page devoted specifically to the router
itself at http://www.live.com/ wireless/unix-base-station.html.
Cisco 1200 Access Point
Cisco's 1200 Series access point is the standard-bearer for standalone access points. It runs a
version of Cisco's Internetwork Operating System (IOS). At first, 1200s ran a system based on
VxWorks, but Cisco released an IOS upgrade to bring the 1200 in to the fold. The upgrade tool is
available from Cisco's support web site. No new features or continuing development has been done
on VxWorks for quite some time.
On an individual basis, the Cisco 1200 can be managed from either a web interface or an IOS
command-line interface. Larger installations can be managed through the Wireless LAN Solutions
Engine (WLSE). The command-line interface is available through a local console serial cable with a
Cisco RJ-45 pinout, or once the network interface is configured, telnet or SSH.
Setting Up the 1200
Hardware setup on the 1200 is straightforward. It may be powered either from a local power supply
or with a Cisco-specific power injector. The power circuitry is the same. The power supply puts out 48
volts, just as a power injector would.
The 1200 uses a Bridge-group Virtual Interface (BVI). BVIs are a software construct used in IOS that
allows routing and bridging of protocols over the same interface. APs need to bridge 802.11 frames
to 802.3, but also need to route IP for management purposes, which makes BVIs the obvious choice.
Configure the IP address on a BVI from the global configuration prompt. APs may get their IP
address either from a static assignment, or from a built-in DHCP client. Both commands are shown
below:
ap1200# configure terminal
ap1200(config)# interface BVI1
ap1200(config-if)# ip address 192.168.1.5 255.255.255.0
ap1200(config-if)# ip address dhcp client-id FastEthernet0
To check on the status of the interface and see what addresses have been assigned, use show ip
interface:
ap1200#show ip interface brief
Interface
IP-Address
Dot11Radio0
unassigned
FastEthernet0
unassigned
Virtual-Dot11Radio0
unassigned
BVI1
192.168.5.191
Configuring Radio Interfaces
OK?
YES
YES
YES
YES
Method
TFTP
NVRAM
TFTP
DHCP
Status
up
up
down
up
Protocol
up
up
down
up
The 1200 has two radios. Radio 0 is the 2.4 GHz radio, which is usually 802.11g, but may be
802.11b in older hardware. Radio 1 is the 5 GHz radio. Each radio can be configured independently
by using the interface configuration commands. Data rates can be allowed simply by entering them
with the speed command, or they may be labeled as required by prefacing them with basic-. In the
following command list, the first speed command allows all data rates. The second requires 1 Mbps
and 2 Mbps operation, but allows 5.5 Mbps and 11 Mbps operation. The last two are special. speed
range allows all, but requires only the slowest speed. speed throughput sets all data rates to
required.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# speed 1.0 2.0 5.5 11.0
ap1200(config-if)# speed basic-1.0 basic-2.0 5.5 11.0
ap1200(config-if)# speed range
ap1200(config-if)# speed throughput
Transmission power may be configured for each radio by setting a local maximum power in
milliwatts. Like speed, power is an interface-specific configuration command. 802.11b/g radios are
capable of up to 100 mW power. Due to limitations in chip design with OFDM, 802.11a radios are
only capable of 40 mW transmission.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# power local 100
In addition to power settings, the operating channel may be configured for each radio with the
channel command. As an argument, channel takes the frequency in MHz. Alternatively, using the
least-congested keyword will force the AP to monitor all channels and pick the clearest one.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# channel 2412
ap1200(config-if)# channel least-congested
Two different types of 802.11b preambles are used. Long preambles are more compatible, but short
preambles give much better performance. Generally, this option should be set to short unless there is
a known older device that needs to use the network. To disable short preamble, use the no
preamble-short interface command.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# no preamble-short
Beacon frames are used to announce the existence of a network, as well as announce buffered with
DTIM information elements. The Beacon interval can be tuned to balance delivery of buffered frames
with battery consumption using the beacon period and beacon dtim-period commands.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# beacon period 100
ap1200(config-if)# beacon dtim-period 5
In addition to the Beacon interval, the RTS/CTS threshold can be configured. Lower values will cause
RTS/CTS handshaking to occur. Depending on the environment, it might also be worth altering the
number of times a frame will be retransmitted, or the threshold at which it will be fragmented using
the following commands.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# rts threshold 2000
ap1200(config-if)# rts retries 2
ap1200(config-if)# packet retries 8
ap1200(config-if)# fragment-threshold 1500
Internetworking
Different clients will encapsulate frames in different ways. By far the most common is to use RFC
1042 SNAP encapsulation, which is the default. IOS allows configuration of 802.1H as well. The
setting is global for a radio, and cannot be configured on a per-protocol basis.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# payload-encapsulation snap
ap1200(config-if)# payload-encapsulation dot1h
With dynamic VLAN assignment, the use of VLANs on the wired side is becoming increasingly
common. IOS on the 1200 supports both native and tagged VLANs. It is important to assign VLAN 1,
the native VLAN, to the same IP network as other devices on the network to ensure communication
across the so-called native VLAN. The native VLAN is noted by adding the keyword native at the end
of an encapsulation command for the subinterface.
ap1200# configure terminal
ap1200(config)# interface dot11radio0.1
ap1200(config-subif)# encapsulation dot1q 1 native
ap1200(config-subif)# interface fastethernet0.1
ap1200(config-subif)# encapsulation dot1q 1 native
Further VLANs can be configured in a similar way, omitting the native. It is common practice to keep
the subinterface number equal to the VLAN tag. For example, to configure VLAN 10, the following
commands would be used. To configure VLAN 20, all the references to 10 would be replaced with 20.
ap1200# configure terminal
ap1200(config)# interface dot11radio0.10
ap1200(config-subif)# encapsulation dot1q 10
ap1200(config-subif)# interface fastethernet0.10
ap1200(config-subif)# encapsulation dot1q 10
Configuring Security
Radio networks are broadcast from an AP as an SSID. Each SSID acts somewhat like its own virtual
self-contained access point within the 1200. Each SSID can have its own security configuration, as
well as its own VLAN mapping. Interestingly enough, the VLAN mapping is slightly spongy. A default
VLAN can be assigned to an SSID. If the RADIUS server in use returns a different VLAN, the client
device will be re-mapped on to the specified VLAN.[] The per-SSID default will only be used when
nothing is supplied by the RADIUS server. (This approach is not quite as clean as many other
products on the market, which work on an either/or basis with default VLANs and RADIUS servers.)
[]
Cisco APs require a set of RADIUS tunnel attributes to be supplied, and they must be tagged. It requires the Tunnel-Type
attribute to be set to "VLAN", the Tunnel-Medium-Type attribute to be set to IEEE-802, and the Tunnel-Private-Group-ID attribute
to be set to the VLAN ID.
Authentication to each SSID is configured using the authentication command. In general, this
command will be set to "open" authentication, but may add EAP authentication as an optional
method. The following commands configure an SSID of babelfish to map to VLAN 42 while requiring
EAP authentication against the server group rad_eap.
ap1200# configure terminal
ap1200(config)# interface dot11radio 0
ap1200(config-if)# ssid babelfish
ap1200(config-ssid)# vlan 42
ap1200(config-ssid)# authentication open eap rad_eap
To define the RADIUS server for EAP authentication, define each server, and associate it with a
group. By default, RADIUS servers that are not assigned UDP ports are assigned to the old RADIUS
ports (1645 and 1646), so they must be explicitly assigned to the new ports.
ap1200# configure terminal
ap1200(config)#radius-server host 192.168.200.187 auth-port 1645 acct-port 1646
key MySecret
ap1200(config)#radius-server host 192.168.200.188 auth-port 1812 acct-port 1813
key MySecret
ap1200(config)#aaa group server radius rad_eap
ap1200(config-sg-radius)#server 192.168.200.187 auth-port 1645 acct-port 1646
ap1200(config-sg-radius)#server 192.168.200.188 auth-port 1812 acct-port 1813
Configuring WPA-PSK
WPA preshared keys are configured through the SSID command. By setting the SSID up for WPA key
management, it is possible to specify either an ASCII or hexadecimal pre-shared key for use with
WPA.
ap1200(config)#interface dot11radio 0
ap1200(config-if)#ssid LuminiferousEther
ap1200(config-ssid)#authentication key-management wpa optional
ap1200(config-ssid)#wpa-psk ascii Thisisaverylongsecretpresharedkey!
Monitoring
A basic monitoring tool is the list of associated stations, which can be obtained from an unprivileged
prompt:
ap1200> show dot11 association
802.11 Client Stations on Dot11Radio0:
SSID [LuminiferousEther] :
MAC Address
IP address
Device
0002.2d6e.abda 192.168.200.150
-
Name
-
Parent
self
State
Assoc
To view details on a particular association, ask for it by MAC address. The complete association
record will be printed out, including the types of encryption in use. This station is associated and uses
TKIP for encryption.
ap1200> show dot11 association 0002.2d6e.abda
Address
: 0002.2d6e.abda
Name
:
IP Address
: 192.168.200.150
Interface
: Dot11Radio 0
Device
:
Software Version :
CCX Version
:
State
: Assoc
SSID
: LuminiferousEther
Hops to Infra
: 1
Clients Associated: 0
Tunnel Address
:0.0.0.0
Key Mgmt type
: WPA PSK
Current Rate
: 11.0
Supported Rates
: 1.0 2.0 5.5 11.0
Signal Strength
: -39 dBm
Signal Quality
: 79 %
Power-save
: Off
Parent
: self
VLAN
: 0
Association Id
: 120
Repeaters associated: 0
Encryption
Capability
: TKIP
:
Connected for
: 1463 seconds
Activity Timeout : 55 seconds
Last Activity
: 4 seconds ago
Packets Input
Bytes Input
Duplicates Rcvd
Decrypt Failed
MIC Failed
MIC Missing
:
:
:
:
:
:
535
61629
0
0
0
0
Packets Output
Bytes Output
Data Retries
RTS Retries
:
:
:
:
245
137018
18
0
Troubleshooting
IOS has extensive debugging facilities that can be used to troubleshoot problems. Tracing is
activated by the debug command, which is followed by the area to perform tracing on. By default,
tracing is sent to the console. If you are attached to the device over the network, you will need to
send the debugging output to the current login screen with the following command:
ap1200# terminal monitor
The most troublesome part of working with secure 802.11 networks is the initial association and key
distribution phase. Troubleshooting of these actions can be accomplished with debug dot11 and its
subcommands. Some common troubleshooting debugging commands are shown in Table 20-2.
Table 20-2. Cisco 1200 debugging commands
Debug area
Commands
Remarks
debug radius authentication
EAP
authentication
debug dot11 aaa authenticator
process
debug dot11 aaa authenticator
state-machine
MAC filtering
debug dot11 aaa authenticator
mac-authen
Prints out RADIUS packets; decodes
attributes; explains actions
May show servers timeout or fail
Shows MAC addresses and response from
authentication system
debug dot11 aaa authenticator
process
WPA
debug dot11 aaa authenticator
state-machine
Shows key exchange
debug dot11 aaa manager keys
Turning off debugging is simple. Just enter the following command:
ap1200# undebug all
Apple AirPort
There are now two types of AirPort base stations that are on the market from Apple. Both the AirPort
Express and AirPort Extreme Base Station are 802.11g devices. The AirPort Extreme is designed
more as a home gateway, and it has a modem and external antenna port. Although more costly than
a number of other devices on the market, it is also more functional and is a common small-office
solution.
First-Time Setup
The first-time setup is done with the same AirPort Setup Assistant application that configures new
wireless interfaces. AirPorts fresh out of the shrink wrap broadcast their existence to the world so
that a Mac OS client can be used for configuration. After connecting to the AirPort and clicking
through to the configuration screen, the network security setup screen of Figure 20-3 is presented.
The configuration displayed shows a WPA pre-shared key security method. It is also possible to
configure WEP from this screen.
Figure 20-3. Network security settings
The Management Interface
Once the bootstrap configuration is done with the Setup Assistant, the AirPort Base Station is on the
network and must be configured with the AirPort Utility (Figure 20-4). This is a separate
configuration utility that will feel vaguely familiar if you have ever seen Lucent's AP Manager. When it
is started, the AirPort Admin Utility searches all the AirPort base stations on the network and displays
them in a list. Individual base stations can be selected for further configuration. When changes are
made, the base station must be restarted for the changes to take effect. The "Other" button at the
top allows configuration of any AirPort Base Station that the manager can send IP packets to. Faraway base stations may not appear on the browse list, but by clicking on "Other" and entering an IP
address, the Manager can configure any base station to which it has IP connectivity.
Figure 20-4. AirPort Admin Utility main screen
Configuring the wireless interface
When a base station is selected for configuration, the configuration screen will pop up. Several tabs
are used to group configuration information into logical subsets, and the wireless interface
configuration is available by default. Across the top, there are buttons to restart the access point,
upload new firmware, return the base station to factory defaults, and change the password. (New
firmware is distributed as part of the Admin Utility package, and will automatically be used to update
old AirPorts as the manager discovers them.)
The AirPort configuration is shown in Figure 20-5. Security settings are similar to those shown in
previously in Figure 20-3. The "closed network" option prevents the inclusion of the SSID information
element in Beacon frames and requires that stations associating provide it. It is a toy security option,
and does not really close the network. As dual-mode 802.11b/g devices, it is possible to configure
them in a b/g compatibility mode, or attempt to keep 802.11b devices off the network to reduce the
overhead of protection.
Figure 20-5. Wireless interface configuration
Configuration of the LAN interface
The AirPort may be used as a NAT device, in which case it will translate IP traffic from wireless
devices on to its address on the Ethernet interface. Client computers can be statically addressed by
the network administrator, or the built-in DHCP server can be turned on and assigned a range of
addresses to assign to clients. When NAT is used to hide several computers behind one IP address,
the address specified in the Internet properties is used as the public address. The wired LAN interface
is given the private address 10.0.1.1, and DHCP is used to lease out addresses from 10.0.1.2 to
10.0.1.200 (in this case, you can't select the range of addresses for the DHCP server to give out).
The AirPort base station can be connected to wired networks by its Fast Ethernet LAN port if the
DHCP server is disabled.
The Port Mapping tab, shown in Figure 20-6, can be used to add inbound static port mappings. The
public port is translated to the private IP address and port number listed in the mapping. The figure
illustrates an address translation for inbound web services to port 80 on host 10.0.1.201.
Figure 20-6. Port Mapping tab
Access control
Like most other products, the AirPort Base Station supports filtering by client MAC address. The
Access Control tab lets you identify clients by their AirPort ID (MAC address) and add them to a list of
allowed clients, together with a description.
Chapter 21. Logical Wireless Network
Architecture
Planning a wireless LAN installation is a significant undertaking that cuts across many previously
disparate disciplines. This chapter begins the discussion of wireless LAN deployment by tackling the
network architecture. Network design is about trade-offs between several factors, including cost,
manageability, availability, and performance. Wireless networks add the additional dimension of
mobility to the mix.
Wireless networks often extend an existing wired infrastructure. The wired infrastructure may be
quite complex to begin with, especially if it spans several buildings in a campus setting. Wireless
networks depend on having a solid, stable, well-designed wired network in place. If the existing
network is not stable, chances are the wireless extension is doomed to instability as well.
This chapter discusses four approaches to building a wireless LAN. All are discussed in terms of the
technical features of wireless LANs that influence how you design your wireless network. How do the
features of wireless LANs influence network topology? Besides the 802.11 equipment, what other
equipment is needed to deploy a network? How should the logical network be constructed for
maximum mobility?
Evaluating a Logical Architecture
Before presenting any topologies in detail, I will discuss how to evaluate a proposed network. Each of
the topologies presented in this chapter has its own strengths and weaknesses. Choosing a topology
depends on which criteria are compelling on your network. Here are discuss several evaluation
criteria that I have found to be important.
Mobility
Portability results in a productivity gain because users can access information resources wherever it
is convenient to do so. At the core, however, portability removes only the physical barriers to
connectivity. It is easy to carry a laptop between several locations, so people do. But portability does
not change the ritual of connecting to networks at each new location. It is still necessary to physically
connect to the network and reestablish network connections, and network connections cannot be
used while the device is being moved.
If, for example, you have a remotely mounted filesystem and put a laptop to sleep, it may not be
there when the laptop wakes up in a different location. Any open files or activity against the
filesystem may need to timeout before you can regain control of the computer. DHCP clients may
impose more pedestrian restrictions. Common DHCP clients attempt to renew the lease on the last
address obtained, and may need to go through multiple DHCP exchanges to get an address on a new
IP subnet.
Mobility, on the other hand, is a far more powerful concept: it removes further barriers, most of
which are based on the logical network architecture. Network connections stay active even while the
device is in motion. This is critical for tasks requiring persistent, long-lived connections, which may
be found in database applications. Support personnel frequently access a tracking database that logs
questions, problems, and resolutions. The same argument can be made for a number of tracking
applications in a health care setting. Accessing the database through a wireless network can boost
productivity because it allows people to add small amounts of information from different locations
without needing to reconnect to the database each time. Inventory applications are another example
and one of the reasons why retail and logistics are two of the markets that have been quicker to
adopt 802.11. When taking inventory, it makes far more sense to count boxes or products where
they sit and relay data over a wireless network than to record data on paper and collate the data at
the end of the process.[*]
[*]
Indeed, the early adopters of wireless LAN technology tended to be in organizations where the work was mobilehealth care,
logistics, and education.
Traditional wired Ethernet connections provide portability. I can take my laptop computer anywhere
on the campus at work and plug in. (If I'll tolerate slow speeds, I can even make a phone call and
access my corporate network anywhere in the world.) Each time I access the network, though, I'm
starting from scratch. I have to reestablish connections, even if I only moved a few feet. What I'd
really like is to walk into the conference room and connect to the corporate network without doing
anything.
Defining "mobility"
Wireless networking and mobility are intertwined concepts. Without mobility, wireless networking
would not be particularly interesting. Mobility means that applications just work, no matter where the
computer is. Unfortunately, building a network that provides location-independent services requires a
great deal of location-based configuration and knowledge.
Translating the high-level definition of mobility into technical details can be done in several different
ways. Many technologies can be used to provide mobility for network users, and not all of them are
good. Providing network transport that is truly independent of the application and transparent to it
has several requirements.
1. Consistent MAC-layer attachment to the same link-layer network. Mobility requires that the
existing data-link layer look the same regardless of location. Extensive engineering in 802.11
provides link-layer mobility.
a. Transparent handoffs between access points. Users may need to perform an initial
configuration operation to select a network to connect to, but they should not be involved
in handoff decisions. If the signal strength gets too low, the software should attempt to
locate a better signal and transfer to it without user intervention. Every wireless card I am
aware of currently switches between access points that are part of the same network.
802.11 was designed around this requirement, and any piece of wireless LAN equipment
you buy should have no trouble meeting it.
b. The inter-access point handoff can break down when the access points are located in
different broadcast domains. If two stations are exchanging frames and one moves to a
new access point, 802.11 cannot guarantee that the two stations stay attached to the
same broadcast domain. One opportunity for vendors to improve on 802.11 is to link
access points with the same SSID together so they no matter where they attach to the
network, stations in the wireless LAN attach the same broadcast domain everywhere, even
if the access points are connected to different local broadcast domains.
c. Moving a client system between two access points also requires either setting up a new set
of security parameters for encryption and integrity protection, or transferring security
parameters from the old access point to the new access point. 802.11 does not specify
either procedure. Depending on the hardware in use, the process of establishing a security
context with the new access point can take a fair amount of time. In bulk data transfer,
the blip may not be noticeable; for voice conversations, it often is.
2. No configuration changes are required to the client network stack. For the most part, this
means that clients can maintain the same network address as they move throughout the
network. In most networks, the address is an IP address, although there may be some
applications that require maintenance of other network protocol addresses as well. Extensive
engineering has been devoted to network layer mobility, especially in the IP world, which
traditionally has not provided it. Address maintenance is a major network engineering
challenge, especially if it is required across what would otherwise be an IP subnet boundary.
a. Before maintaining an address, the client must obtain one. At the initial connection to a
wireless LAN, the system should somehow get an address, most likely through DHCP. That
address should be used for the duration of the wireless LAN session. From the client's
perspective, it keeps the IP address as it moves through the network, and does not need
to alter its address or any other stack information to attach to any new access points.
b.
b. Wireless networks can have brief interruptions of connectivity as clients move between
coverage areas or suffer from transient radio link problems. For brief interruptions, access
points should maintain enough session information so that the client can simply rejoin the
network. Access points may also need to buffer frames so that clients can fetch any traffic
that was received during the interruption.
c. Equally important, the client must appear to maintain its address. If the client connects to
a server somewhere, the server should be able to use the same address for the duration of
the connection. Furthermore, any other state the network maintains is often associated
with an IP address. Many networks use NAT to reach the Internet, and the NAT records are
associated with the client IP address.
d. Depending on the application, it may be necessary to have logical network path
preservation. Transmitted frames must always emerge from the same egress point on the
network, so they are subject to whatever controls the network requires. No matter what
the location of the client throughout the network, it always appears as if it is attached to
the same point at the edge of the network.
No single technology supplies all of the components of mobility. At the link layer, much is provided
by 802.11, although additional functionality is often required above and beyond what the standard
lays out. By and large, moving the association between access points is easy. In networks that do not
use link-layer security, bridging records can be transferred between access points in a few
milliseconds to tens of milliseconds. Re-establishing the link-layer security context may take a few
hundred milliseconds, and depends heavily on the responsiveness of the authentication server. Some
companies have devoted significant engineering resources to building products that accelerate the
security context establishment. Newer "Wi-Fi switch" products may also speed up the roaming
process by holding client association records in a centralized location so that there is no need to
transfer it between APs.
At the network layer, several different approaches may be taken to offer network-layer mobility.
Several early devices used address translation (NAT). NAT is not, never has been, and never will be a
mobility protocol. By dragging the network infrastructure up into higher protocol layers, any failure of
the translation breaks applications. Some applications are almost incompatible with NAT, such as
H.323, while others require specific application support, such as IPsec's use of NAT Traversal. Some
devices rely heavily on NAT to provide mobility. Avoid them.
Some form of tunneling is often used to provide application-independent mobility. Client devices are
designated with a "home" network, and then inter-AP protocols automatically direct traffic back to
the home location from any place it is not directly accessible. Tunneling protocols must be defined at
the network layer so that tunnels can carry data across arbitrary network boundaries. However, the
tunneling protocol itself may work at the link layer (making VLAN attachment points available
throughout the network) or at the network layer (making IP addresses routable throughout the
network). The only open industry standard tunneling protocol is Mobile IP (see sidebar later in this
chapter).
One challenge network architects often face is the distance over which mobility must be provided.
Small installations are easy because there are a number of techniques that offer comparable
functionality for up to 25 access points. The real challenges come when designing mobility for larger
installations, or organizations that have widely distributed locations. Part of the key to designing a
successful mobility solution is to determine what users expect. What sort of tasks require that an IP
address be continuously maintained? Interactive terminal sessions, such as telnet and SSH, require
that the source IP address remain the same. Other applications may be able to initiate new
connections upon reattaching to the network, and the process of reconnecting may be transparent to
the users. Generally speaking, most users do not expect to maintain an IP address if they need to
use a car, train, or plane to travel between network sites. Still, large, spread-out campuses might
require higher-than-average mobility support.
Call it the "dinosaur juice" rule for mobility: if you burn fossil fuels for
transportation, it is probably acceptable for the IP address to change.
Mobile IP
One of the most talked-about developments in wireless networks in the past few years is
the emergence of the new "Wi-Fi switch" architecture. In broad terms, Wi-Fi switches
move network functions from access points into an aggregation device that provides
control and management functions.
802.11 performs a sleight-of-hand trick with MAC addresses: stations communicate with
a MAC address as if it were fixed in place, just like any other Ethernet station. Instead of
being fixed in a set location, however, access points note when the mobile station is
nearby and relay frames from the wired network to it over the airwaves. It does not
matter which access point the mobile station associates with because the appropriate
access point performs the relay function. The station on the wired network can
communicate with the mobile station as if it were directly attached to the wire.
Mobile IP performs a similar trick with IP addresses. The outside world uses a single IP
address that appears to remain in a fixed location, called the home location. Rather than
being serviced by a user's system, however, the IP address at the home location (the
home address) is serviced by what is called the home agent. Like the access point, the
home agent is responsible for keeping track of the current location of the mobile node.
When the mobile node is "at home," packets can simply be delivered directly to it. If the
mobile node attaches to a different network (called a foreign network or visited
network), it registers its so-called foreign location with the home agent so that the home
agent can redirect all traffic from the home address to the mobile node on the foreign
network.
Consider two wireless LANs built on different IP subnets. On its home subnet, a wireless
station can send and receive traffic "normally."
When the wireless station moves from its home subnet to the second ("foreign") subnet,
it attaches to the network using the normal procedure. It associates with an access point
and probably requests an IP address using DHCP. On a wireless station that is unable to
use Mobile IP, connections are interrupted at this point because the IP address changes
suddenly, invalidating the state of all open TCP connections.
Wireless stations equipped with Mobile IP software, however, can preserve connection
state by registering with the home agent. The home agent can accept packets for the
mobile station, check its registration tables, and then send the packets to the mobile
station at its current location. The mobile station has, in effect, two addresses. It has its
home address, and it can continue to use this address for connections that were
established using the home address. It may also use the address it has been assigned on
the foreign network. No TCP state is invalidated because the mobile station never
stopped using its home address.
Naturally, system administrators (rightly) rebel at installing new software on end-user
systems. An alternative is to use proxy mobile IP, in which the network edge
incorporates the functions of the agent software. Moving Mobile IP functions into the
access points eliminates the need to install client software, although it does increase the
complexity of the access points.
I have omitted a great deal of the protocol operations. Designing a protocol to allow a station to
attach anywhere in the world and use an address from its home network is a significant engineering
endeavor. Several security problems are evident, most notably the authentication of protocol
operations and the security of the redirected packets from the home network to the mobile station's
current location. Maintaining accurate routing information, both the traditional forwarding tables at
Internet gateways and the Mobile IP agents, is a major challenge. And, of course, the protocol must
work with both IPv4 and IPv6. For a far more detailed treatment of Mobile IP, I highly recommend
Mobile IP: Design Principles and Practices by Charles Perkins (Prentice Hall).
Security
As the networks of the world have united into a single, globe-spanning behemoth, security has taken
on new importance. Wireless LANs were once the bane of security-conscious networking
organizations, but newer tools make it easier to build networks with significant security protections.
In addition to traditional security issues such as traffic separation between user groups and
maintaining appropriate access privileges, wireless networks present new challenges, like rogue
access points and unauthorized clients.
Many of logical architecture's security ramifications are related to the selection of encryption and
authentication protocols. But the architecture may have some additional security implications based
on technology available at the edge of the network.
Traditional access points are autonomous devices that act as independent network elements. Access
points are placed out in user areas, and are usually not physically secured. Unfortunately, the
popularity of 802.11 may make unsecured access points in public areas targets for theft. Traditional
access points have local software and configuration, and may also be attractive targets to attackers
who can use learn sensitive security information from the configuration, such as the RADIUS shared
secret. Newer "thin" access points help address this concern by removing a great deal of functionality
from the access point and pulling it back into secured controllers that can be locked away in wiring
closets.
Attackers may also remove access points to obtain additional network privileges. In some
architectures, access points must be connected to relatively privileged network ports. If, for example,
several VLANs are made available through a wireless network, many access points require that the
access point connect to a link tagged with all the available VLANs. In such a setup, an attacker may
obtain direct access to the backbone by removing an access point and connecting to its port.
One additional concern is that many government networks need to be designed to comply with
regulations and best practices. The National Institute of Standards and Technology (NIST) is
responsible for developing many of the computing standards used by the U.S. federal government;
these standards are referred to as Federal Information Processing Standards (FIPS). One particular
standard, FIPS-140, lays out requirements for secure network designs. Not surprisingly, FIPS-140
specifies that certain types of data must be encrypted. Less obviously, FIPS-140 also requires that
approved encryption algorithms and modes must be used. Not all security protocols and standards
are created equal, and only some network designs are capable of meeting FIPS-140 criteria. FIPS
requirements affect most federal agencies, but may also exert a powerful influence by requiring
compliance on the part of an agency's suppliers and contractors. (FIPS requirements are discussed in
more detail in Chapter 22.)
Performance
For the benefit of mobility, wireless networks impose a cost. Simply, performance is nowhere near
what can be expected from a well-engineered wired LAN. Wireless networks have smaller advertised
bit rates than wired LANs. To make matters worse, the big number in the glossy brochure omits a
great deal of protocol overhead.[] Table 21-1 gives rule-of-thumb estimates for the maximum
throughput based on the technology.
[]
The effective throughput is actually much lower. Quantifying the exact impact of the protocol overhead is very difficult; for one
straw-man calculation, see http://www.oreillynet.com/pub/a/wireless/2003/08/08/wireless_throughput.html.
Table 21-1. Maximum throughput for different 802.11 technologies
Advertised
throughput
Estimated maximum Estimated maximum
continuous
perceived throughput with
throughput
5:1 multiplexing factor
802.11b
11 Mbps
6 Mbps
30 Mbps
802.11a
54 Mbps
30 Mbps
150 Mbps
802.11g, no protection
54 Mbps
30 Mbps
150 Mbps
802.11g, with
protection
54 Mbps
15 Mbps
225 Mbps
802.11a+802.11b
11+54 Mbps
36 Mbps
180 Mbps
802.11a+802.11g
(without protection)
54+54 Mbps
60 Mbps
300 Mbps
Technology
Single radio systems
Dual radio systems
Unlike wired networks, the speed at which wireless networks operate depends on the distance from
the nearest network uplink. As stations move farther from the serving access point, signals become
weaker and the operational speed falls. Throughput depends on distance from the access point and
the position of the device in a way that is unfamiliar to users of wired networks.
Throughputs in the neighborhood of Table 21-1 require very strong signals, and by necessity, limited
distances from the access point. As a practical matter, most wireless LANs have been built for
coverage. At the fringes of connectivity, throughput is at the lowest rate for the technology, not the
highest rate. Figure 21-1 illustrates the typical scenario. Although stations close-in connect at the
maximum data rate, stations that are further away require much longer time intervals to transmit
the same data. Even if the radio medium is in constant use, the maximum throughput of the stations
associated to the AP in Figure 21-1 would be much less than 6 Mbps.
Even though the maximum data rates shown in Figure 21-1 are small, the magic of statistical
multiplexing can make the rate seem much bigger than it is. Networks are based on the principle that
user traffic is often quite bursty. Although users demand a peak data rate of a megabit per second,
users are often idle. The network only needs to supply the peak data rate during email fetches and
web page downloads; when the connection is idle, other users can get their megabit per second
service. In my experience, a multiplexing factor of between 3:1 and 7:1 can be used to account for
the burstiness of network traffic. Although an 802.11b network may only deliver 6 Mbps, it can
accommodate 20-30 users demanding megabit service because very few applications require
continuous service. (Voice applications are a notable exception.) The last column in Table 21-1 shows
the effective throughput with a 5:1 multiplexing factor. Depending on applications and experience,
you may need to use a higher or lower factor on your network.
Figure 21-1 is only a qualitative display of how the speed of a network changes with ditsance. Freespace loss calcuations can be used to get a more quantitative grasp of a network's range and
propagation characteristics. Free-space loss, or path loss, is the transmission loss over space, with
no obstructions. It is an "ideal" number, describing only how the electromagnetic waves diminish in
strength over distance while neglecting a number of factors that are relevant to building 802.11
networks. Indoor networks must deal with walls, windows, doors, and the fact that many signals may
not be line of sight. Free-space loss calculations assume the noise floor is also sufficiently low that
the receiver sensitivity is the limiting factor. In many environments, the number of unlicensed
devices operating in the 2.4 GHz ISM band push up the noise floor well past the lower limits of many
receivers.
Figure 21-1. Dependence of performance on range
Free-space loss depends only on two inputs: the frequency of the signal and the distance it must
cover. Higher frequencies attenuate faster, as do longer distances.
Table 21-2 is based on a free-space calcuation for both 2.4 GHz and 5 GHz wireless LANs. For the
ISM frequency, it uses 2.437 GHz, the center frequency of channel 6. For the 5 GHz frequency, it
uses 5.250 GHz, which is the midpoint of the U-NII lower and mid-bands. (Some early 802.11 cards
were only able to function in the two low bands, so some networks restrict 802.11a channel usage to
the first 8.) It assumes that power is transmitted at the maximum of 20 dBm (100 mW) in the 2.4
GHz band, and at the maximum practical power of 14 dBm (25 mW) for many 802.11a devices. For a
receiver cut-off, the sensitivity listed on the data sheet for the Cisco CB-21 a/b/g client card was
used. By using the sensitivity directly, it is an implied assumption that the noise floor is significantly
below the values listed in the sensitivity column. Do not expect to get the ranges noted in this table,
since it is only a calculation of the maximum ideal range. It is, however, a useful guide to comparing
different modulation rates as well as the different frequency bands in use.
Table 21-2. Free-space range for different modulation speeds
Modulation
Sensitivity
type and speed (Cisco CB-21)
Maximum freespace range
(meters)
Free-space range
relative to
maximum speed
Percentage of
maximum range
modulation
2.4 GHz
1 Mbps DSSS
-94
4,850
13.9
100%
2 Mbps DSSS
-93
4,300
12.3
89%
5.5 Mbps CCK
-92
3,850
11.0
79%
11 Mbps CCK
-90
3,050
8.7
63%
6 Mbps OFDM
-86
1,930
5.5
40%
9 Mbps OFDM
-86
1,930
5.5
40%
12 Mbps OFDM
-86
1,930
5.5
40%
18 Mbps OFDM
-86
1,930
5.5
40%
24 Mbps OFDM
-84
1,530
4.4
32%
36 Mbps OFDM
-80
970
2.8
20%
48 Mbps OFDM
-75
550
1.6
11%
54 Mbps OFDM
-71
350
1.0
7%
6 Mbps OFDM
-89
630
7.0
100%
9 Mbps OFDM
-89
630
7.0
100%
12 Mbps OFDM
-89
630
7.0
100%
18 Mbps OFDM
-85
400
4.4
63%
24 Mbps OFDM
-82
280
3.1
44%
36 Mbps OFDM
-79
200
2.2
32%
48 Mbps OFDM
-74
110
1.2
18%
54 Mbps OFDM
-72
90
1.0
14%
5 GHz
The performance of a wireless network differs from a wired network in another important way. Most
wireline networks are built using switches, so multiple independent traffic flows can be switched
independently. Wireless networks are inherently a shared medium. Within an access point's coverage
area, only one client may transmit data. Wireless networks often are built simply to cover an area. As
they become popular, network administrators must add capacity by shrinking the coverage area of
individual APs to reduce bandwidth contention. Limiting bandwidth contention by making coverage
areas smaller is the approach taken by most "Wi-Fi switches."
Physical characteristics of the wireless medium negatively affect performance and service quality.
802.11 uses a fully-acknowledged MAC for reliability, but the need for acknowledgments increases
the transmission latency for frames. Lost packets and the potential for retransmission may also
increase the variability in latency, called jitter. For some applications, the difference in service is
imperceptible to the user. Bulk data transfers depend much more on available capacity than the
service quality. However, if you plan to run voice on your wireless network, you will need to spend
some time engineering acceptable service quality. Voice depends on regular delivery of small
amounts of data, not high throughput. Some of the architectures discussed in this chapter are better
suited to applications with demands for high quality network service.
Traditional quality of service queuing and controls are important not only because of the need to
cope with the relatively low reliability of wireless signals, but also because the bandwidth is so
limited. With only a few megabits per second on each access point, a single ill-behaved client can
soak up all the available capacity. Some access points provide better traffic shaping and policing to
engineer allocation of scarce wireless medium capacity than others, and some logical architectures
are better able to cope with high-bandwidth applications.
As with many other network engineering tasks, building a wireless LAN with the necessary
performance is a matter of analyzing the types of applications you intend to run on the LAN and
removing as many bottlenecks as possible. In some cases, it may be possible to build enough
network capacity to provide the required user experience. For other applications, it may be necessary
to manage resource allocations using traffic management tools on the access points making up the
network.
Backbone Engineering
Access points are the edge of the ever-expanding network. In most cases, they do not offer new
network services; they just make existing services more widely available and easy to use. As edge
devices, though, they must plug into a network core to interconnect users with resources. Expect to
do some additional configuration of the network core when wireless is added to the edge of the LAN.
When 802.11 was a new standard and products had just emerged, there was no mobility except for
what the network engineers created, and no traffic separation at all. Mobility was entirely up to the
network engineer. If the access points could be placed in a single link layer network, mobility existed.
If not, mobility remained elusive. Early wireless LANs required, as a matter of practice, that the
network be built around a switched core so that a single subnet snaking through the entire campus
could be used to connect up all the access points. Many newer networks are built around a switched
core, but older networks frequently have constraints that prevents building a purely switched core.
After the first wave of products arrived, the industry produced devices that could attach users to
multiple VLANs. Although these products allowed for multiple user groups over the air and the
backbone wire, the backbone configuration impact was even larger. Rather than extending a single
VLAN through the campus for the access points, every VLAN had to be extended to every access
point throughout the campus. One of the goals of centralized Wi-Fi controllers is to further reduce
backbone reengineering by allowing for simpler configurations of the access device connections.
Beacons, BSSIDs, and VLAN integration
Extending existing VLANs over the air is one of the major tasks that standards groups are working
on. In the meantime, vendors have taken a variety of approaches. As you evaluate these
approaches, there are a few fundamentals.
802.11 built in the concept of having multiple "service sets," but did not explicitly define what a
service set was. Moreover, there are two types of service set identifiers (SSIDs) in 802.11. Extended
Service Set IDs (ESSIDs) are "network names."[*] Multiple APs can be configured to advertise a set
of connections into the air. When a client wishes to connect to a wireless network, it issues probes for
the ESSID it is trying to find, and APs belonging to that SSID respond. ESSIDs can be transmitted in
Beacon frames, although they do not have to be. (ESSID hiding is a security-through-obscurity
practice; if you flip back to Chapter 9 you will note that the Probe Response frame includes the
unencrypted SSID.) The second type of service set, the Basic Service Set ID (BSSID), is the MAC
address of the AP. It is used as the transmitter or receiver address on frames that are bridged
between the wireless and wired networks.
[*]
This is a bit of a simplification. ESSIDs are only network names for infrastructure networks. As always, this book generally
assumes that you are running an infrastructure network, not an ad hoc network.
Generally speaking, each ESSID being transmitted should have its own BSSID. Although there is
nothing to prevent an access point from transmitting multiple ESSIDs in a single Beacon, or even
responding to ESSIDs not included in its Beacon, such behavior causes problems for many drivers.
The Windows Zero Configuration software, for example, generally does not accept a configuration for
a secondary, hidden ESSID. Rather than probing for the hidden ESSID, it attempts to attach to the
transmitted ESSID in the Beacon.
When APs were first able to attach users to different VLANs, a common way of providing the
configuration to users was to expose each VLAN as an SSID, and allow users to choose. While such
an approach allows a great deal of flexibility in the way that users are assigned to VLANs, and allows
them to switch between VLANs as circumstances dictate, the flexibility is a double-edged sword that
may lead to confusion for users as they select the wrong configuration. Transmitting Beacon frames
for a number of extra SSIDs requires the use of scarce wireless network capacity. For many
environments, it makes more sense to dynamically assign users to VLANs based on a user profile
database.
IP addressing
IP addresses are often a reflection of the physical network topology, and wireless networks are no
different. Assigning IP addresses is a subordinate decision to the topology that you choose. Some
network designs may require new address assignments and routing configuration, while others do
not. Organizations that attempt to allocate address space hierarchically, perhaps for reasons of
routing table size, may find it difficult to reconcile logical address assignment with the underlying
physical topology. Furthermore, organizations that make use of registered IP address space may find
that it is at a premium.
Network Services
Ideally, new network devices should plug into existing infrastructure and just work with a minimum
of reconfiguration. A few network services are important to the user experience, and should be
considered in some detail.
DHCP
Users expect to be able to plug in to a network and just have it work. Wireless networks are no
exception, and can make the problem worse. Users expect to be able to attach to any wireless
network and have it just work. Practically speaking, the only way to have the IP stack for a wireless
interface self-configure appropriately for an arbitrary number of networks is to use DHCP.
Many access points have built-in DHCP servers. Smaller networks consisting of an access point or two
may choose to use the built-in DHCP service, but larger networks should have a single source of
DHCP addressing information. Access point DHCP servers may not cope with the load very well, and
some access points may reclaim leased addresses when an association lapses.
Any DHCP service should have a view of the total available address space, and this is best done
outside the access point. Access points act like bridges, and pass DHCP requests from the wireless
network through to the wired network, where a DHCP server can reply. With judicious use of DHCP
helpers, a single server can support multiple IP networks, whether they are implemented as VLANs or
multiple wireless subnets.
Users expect DHCP, so use it. For maximum effect, deploy as few DHCP servers
as possible.
Operating system login
One of the challenges to wireless networking is that validating user credentials often depends on
making a network connection to an authentication server. For example, Windows logins are validated
against domain controllers. In a wired world, making that connection is trivial because the network is
available whenever the cable is plugged in. In a wireless network, though, there is a bit of a chicken
and egg problem. User credentials are used to authenticate the network connection, but a network
connection is required to validate the user credentials. Not all operating system vendors have
considered this problem. If yours has not, you may need to configure special login features or use
additional client software to plug the gap.
Client Integration
Different logical architectures require different client software support. It is easy to build a network
that has no client integration beyond the new drivers by leaving out security entirely. However,
security is a requirement for nearly every network.
The most basic level of security is a static WEP key. Some older devices may not support anything
better than static WEP, in which case you are stuck with it in the absence of a forklift upgrade. Static
WEP configuration is typically built into driver software or client configuration utilities, and requires
no more client integration than using the card itself. For security reasons, though, you want to run
something other than static WEP if at all possible.
Link-layer solutions based on 802.1X offer significant advances in security, but require more
sigificant client integration work. 802.1X supplicant software must be loaded on the client system
and set up correctly. Supplicant software is incorporated into recent operating systems (Windows
2000, Windows XP, and Mac OS X 10.3). Older systems may require a client software package to
implement 802.1X.
VPN solutions vary, depending on both the vendor and the type of VPN technology. SSL-based VPNs
work by directing client access through a secure web site that acts as a portal for applications.
Because the technology is based on a secure web site, there is no client installation beyond the web
browser that probably is already installed on most systems. One downside to SSL VPN technology,
however, is securing non-web applications may require additional effort and user retraining. IPsec
VPNs require a much more disruptive software installation, but are much better at handling an
arbitrary IP-based application. One of the reasons that VPN client software loads are more complex is
that most vendors require their own VPN client for maximum functionality between the client and
gateway.
Some organizations may be limited in their ability to impose client software on users. Universities, for
example, often have users who must open VPN tunnels to external network sites. Professional
development classes are often taken by students who have careers, and may need to connect to
corporate resources from the classroom. Generally speaking, IPsec tunnels inside IPsec tunnels with
multiple clients is a configuration that does not work well, if at all.
Topology Examples
After deciding what is important, you can sketch out what the wireless LAN will look like. Broadly
speaking, there are two major ways of deploying a wireless LAN, and the choice depends broadly on
whether you decide to use security at the link layer. This section describes and analyzes four
different major architectures for wireless LANs. To a certain extent, this section presents four fairly
rigid examples. As the market for wireless LAN network hardware matures, equipment may
incorporate features from multiple topologies, allowing you to mix and match the features that best
suit your needs.
Topology 1: The Monolithic Single-Subnet Network
In the beginning, there was one topology. Access points were simple bridges, and served only to
attach wireless stations to the single wired network they were connected to. Without much
networking intelligence in the access point, wireless networks needed to be designed around the
trivial bridging engines in access points. Networks that supported mobility were correspondingly
simple. When access points are simple bridges without any sophisticated knowledge of, say, VLANs
or routing, they must all attach to the same IP subnet. As long as a station stays on the same IP
subnet, it does not need to reinitialize its networking stack and can keep its TCP connections open.
Equipment limitations dictated the resulting network architecture. Every AP was attached to a single
network. While the network provided mobility, it was often difficult to build, especially on large
campuses. In addition to modifying backbone network configuration, administrators had to set aside
new IP address ranges and route appropriately. The architecture was developed to shield wired
networks from the danger of wireless networks in the time before the development of strong security
protocols. These days, the high configuration overhead and management cost of building two parallel
networks has driven this topology nearly to extinction on any network larger than a few access
points.
Figure 21-2 shows the typical early wireless LAN deployment topology. All the APs are connected to a
single monolithic network. The network is a single link-layer domain, and every station connected to
the network is given an IP address on the IP subnet. For this reason, the monolithic architecture may
be referred to as the single-subnet wireless LAN, the walled garden architecture, or occasionally the
VPN architecture. (It should also be noted that most home networks take the single subnet approach,
although typically with only one access point.)[*] The guiding principle of Figure 21-2 is that the
access points in use cannot provide any services other than link-layer mobility, so they must all be
connected to the same logical link layer. Other design decisions underlying this topology help
augment the access control of the wireless device and lower management overhead by taking
advantage of existing services, each of which will be considered in turn.
[*]
Very large homes may require multiple APs. Generally speaking, an AP should be good for coverage over 3,000-5,000 square
feet, which is sufficient for all but the largest homes.
Figure 21-2. The single subnet wireless LAN deployment topology
Mobility
In Figure 21-2, the network linking all the access points, which is often called the access point
backbone, is a single IP subnet. To allow users to roam between access points, the network should
be a single IP subnet, even if it spans multiple locations, because IP does not allow for network-layer
mobility. (Mobile IP is the exception to this rule; see the sidebar earlier in this chapter.) Networklayer mobility is supplied by the use of a switching infrastructure that supports linking all the access
points together, and an IP addressing scheme that does not require anything beyond link-layer
mobility.
In Figure 21-2, the backbone network may be physically large, but it is constrained by the
requirement that all access points connect directly to the backbone router (and each other) at the
link layer. 802.11 hosts can move within the last network freely, but IP, as it is currently deployed,
provides no way to move across subnet boundaries. To the IP-based hosts of the outside world, the
VPN/access control boxes of Figure 21-2 are the last-hop routers. To get to an 802.11 wireless
station with an IP address on the wireless network, simply go through the IP router to that network.
It doesn't matter whether a wireless station is connected to the first or third access point because it
is reachable through the last-hop router. As far as the outside world can tell, the wireless station
might as well be a workstation connected to an Ethernet.
If it leaves the subnet, though, it needs to get a IP new address and reestablish any open
connections. The purpose of the design in Figure 21-2 is to assign a single IP subnet to the wireless
stations and allow them to move freely between access points. Multiple subnets are not forbidden,
but if you have different IP subnets, seamless mobility between subnets is not possible.
Older access points that cooperate in providing mobility need to be connected to each other at layer
2. One method of doing this, shown in Figure 21-3 (a), builds the wireless infrastructure of Figure 212 in parallel to the existing wired infrastructure. Access points are supported by a separate set of
switches, cables, and uplinks in the core network. Virtual LANs (VLANs) can be employed to cut down
on the required physical infrastructure, as in Figure 21-3 (b). Rather than acting as a simple layer-2
repeater, the switch in Figure 21-3 (b) can logically divide its ports into multiple layer-2 networks.
The access points can be placed on a separate VLAN from the existing wired stations, and the
"wireless VLAN" can be given its own IP subnet. Frames leaving the switch for the network core are
tagged with the VLAN number to keep them logically distinct and may be sent to different
destinations based on the tag. Multiple subnets can be run over the same uplink because the VLAN
tag allows frames to be logically separated. Incoming frames for the wired networks are tagged with
one VLAN identifier, and frames for the wireless VLAN are tagged with a different VLAN identifier.
Frames are sent only to ports on the switch that are part of the same VLAN, so incoming frames
tagged with the wireless VLAN are delivered only to the access points.
By making the access point backbone a VLAN, it can span long distances. VLAN-aware switches can
be connected to each other, and the tagged link can be used to join multiple physical locations into a
single logical network. In Figure 21-4, two switches are connected by a tagged link, and all four
access points are assigned to the same VLAN. The four access points can be put on the same IP
subnet and act as if they are connected to a single hub. The tagged link allows the two switches to be
separated, and the distance can depend on the technology. By using fiber-optic links, VLANs can be
made to go between buildings, so a single IP subnet can be extended across as many buildings as
necessary.
Figure 21-3. Physical topologies for 802.11 network deployment
Figure 21-4. Using VLANs to span multiple switches
Tagged links can vary widely in cost and complexity. To connect different physical locations in one
building, you can use a regular copper Ethernet cable. To connect two buildings together, fiber-optic
cable is a must. Different buildings are usually at different voltage levels relative to each other.
Connecting two buildings with a conductor such as copper would enable current to flow between (and
possibly through) the two Ethernet switches, resulting in expensive damage. Fiber-optic cable does
not conduct electricity and does not pick up electrical noise in the outdoor environment, which is a
particular concern during electrical storms. Fiber also has the added benefit of high speeds for longdistance transmissions. If several Fast Ethernet devices are connected to a switch, the uplink is a
bottleneck if it is only a Fast Ethernet interface. For best results on larger networks, uplinks are
typically Gigabit Ethernet.
For very large organizations with very large budgets, uplinks do not need to be Ethernet. One
company I have worked with uses a metro-area ATM cloud to connect buildings throughout a city at
the link layer. With appropriate translations between Ethernet and ATM, such a service can be used
as a trunk between switches.
Address assignment through DHCP
Within the context of Figure 21-2, there are two places to put a DHCP server. One is on the access
point backbone subnet itself. A standalone DHCP server would be responsible for the addresses
available for wireless stations on the wireless subnet. Each subnet would require a DHCP server as
part of the rollout. Alternatively, most devices capable of routing also include DHCP relay. The
security device shown in Figure 21-2 includes routing capabilities, and some firewalls and VPN
devices include DHCP relay. With DHCP relay, requests from the wireless network are bridged to the
access point backbone by the access point and then further relayed by the access controller to the
main corporate DHCP server. If your organization centralizes address assignment with DHCP, take
advantage of the established, reliable DHCP service by using DHCP relay. One drawback to DHCP
relay is that the relay process requires additional time and not all clients will wait patiently, so DHCP
relay may not be an option.
Static addressing is acceptable, of course. The drawback to static addressing is that more addresses
are required because all users, active or not, are using an address. To minimize end-user
configuration, it is worth considering using DHCP to assign fixed addresses to MAC addresses.
As a final point, there may be an interaction between address assignment and security. If VPN
solutions are deployed, it is possible to use RFC 1918 (private) address space for the infrastructure.
DHCP servers could hand out private addresses that enable nodes to reach the VPN servers, and the
VPN servers hand out routable addresses once VPN authentication succeeds.
Security
This is the oldest of the architectures in this chapter, and pre-dates all the work done on link-layer
security in the past several years. It is generally used on networks where link-layer security is not a
priority, either because security is secondary to providing services (as in the case of an ISP) or
because security is provided through higher-layer protocols with VPN technology. Security trade-offs
in wireless network design are discussed in more detail in Chapter 22.
Backbone engineering
Depending on the existing backbone, using this topology may require prohibitive work on the
backbone, or it may be relatively easy. For maximum mobility