Scaling Workgroup Performance with Switching and Fast Ethernet

Scaling Workgroup Performance with Switching and Fast Ethernet
This Technology Guide is one
of a series of guides, published
by ATG, designed to put complex
communications and networking
technology concepts into practical
and understandable terms.
Each guide provides objective,
non-biased information to assist in
Scaling Workgroup
Performance with
Switching and
Fast Ethernet
the internal education, evaluation
and decision making process.
This Technology Guide, as well
as the other Communications and
Networking Technology Guides
in the series, are available
on ATG‘s Web Site.
http://www.techguide.com
Produced and Published by
One Apple Hill, Suite 216, Natick, MA 01760
Tel: (508) 651-1155 Fax: (508) 651-1171 E-mail: [email protected]
ATG’s Communications &
Networking Technology
Guide Series
Table of Contents
Introduction ................................................................2
Workgroup Evolution..................................................3
The Need for a Scalable Migration Capability ..........4
Technology Options....................................................6
Planning and Provisioning the Network ..................13
The Bottom Line ......................................................21
Glossary ....................................................................23
About the Editor…
Gerald P. Ryan is the founder of Connections Telecommunications Inc., a Massachusetts-based company specializing in
consulting, education and software tools which address Wide
Area Network issues. Mr. Ryan has developed and taught
numerous courses in network analysis and design for carriers,
government agencies and private industry. Connections has
provided consulting support in the areas of WAN network
design, negotiation with carriers for contract pricing and services, technology acquisition, customized software development for network administration, billing and auditing of
telecommunications expenses, project management, and RFP
generation. Mr. Ryan is a member of the IDG World
Expo/ComNet steering committee.
This book is the property of The Applied Technologies Group and is
made available upon these terms and conditions. The Applied Technologies Group reserves all rights herein. Reproduction in whole or in part
of this book is only permitted with the written consent of The Applied
Technologies Group. This report shall be treated at all times as a proprietary document for internal use only. This book may not be duplicated in
any way, except in the form of brief excerpts or quotations for the purpose of review. In addition, the information contained herein may not be
duplicated in other books, databases or any other medium. Making
copies of this book, or any portion for any purpose other than your own,
is a violation of United States Copyright Laws. The information contained in this report is believed to be reliable but cannot be guaranteed to
be complete or correct.
Copyright © 1996 by The Applied Technologies Group, One Apple Hill,
Suite 316, Natick, MA 01760, Tel: (508) 651-1155, Fax: (508) 651-1171
E-mail: [email protected] Web Site: http://www.techguide.com
This Pocket Guide examines the growing demand
for scalable, high bandwidth Ethernet solutions in today’s
workgroups. It explains the factors driving the need for
high bandwidth capabilities at the workgroup. The key
technologies under consideration to meet these needs,
Switched Ethernet and Fast Ethernet, are discussed. The
environment in which each of these complementary
technologies should be deployed is also examined. The
Guide looks at key migration and deployment models to
high bandwidth workgroups, based on existing infrastructure and traffic patterns and future network requirements. The reader will also have a usable guide to a scalable solution which will allow the user to deploy the right
solution at the right time, for the right price.
Introduction
Only 15 years ago, 10Mbps was overkill for existing
applications. Network design was relatively straightforward and corporate-wide connectivity was a rare occurrence. As the corporate network has evolved, one key
component to its success has been the deployment of
local area networks (LANs) based on the Ethernet standard. This has enabled businesses to cost effectively interconnect the great majority of their corporate workers on
shared, 10Mbps LANs. Virtually everyone is “on-line”
and businesses are viewing their network as a critical corporate resource. However, several changes are resulting in
traditional shared 10Mbps LANs becoming the bottleneck to network client performance and, therefore,
enhanced corporate productivity.
Workgroup Evolution
Today’s Challenge
Tthe pervasive use of computers in the business
world, the rapid increase in technological power at the
desktop, and the applications developed to utilize that
power has created the need for increased network capacity to the desktop.
As the number of network users grows within the
corporate structure, and as the volume of client/server
traffic increases at a rapid pace, network managers are
faced with the challenge of supporting user expectations
for more services and higher levels of performance.
These services are also being demanded at a lower risk,
due to the central role being played by networks in
today’s corporation.
The new information intensive applications on the
LAN follow new models and profiles which demand fast
response and high bandwidth. They include graphical
applications such as multimedia, CAD/CAM, document
and scientific imaging, process visualization and interactive video. They also include centralized database applications running on powerful servers. On top of these
applications are the newly emerging groupware applications such as Lotus Notes™. Even video applications are
beginning to show up on previously data-only networks.
These applications, combined with the traditional on
line transaction processing, E-Mail, business productivity
tools, and much more, are beginning to create some serious bottlenecks as well as demands for expanded LAN
capacity and faster network performance.
Islands of Traffic Intensity
This traffic explosion, however, is not uniform. New
network arrangements have resulted in focused, high
powered workgroups operating alongside existing workgroups. This is creating islands of traffic intensity. These
islands, or high-powered workgroups, are emerging with2 • Scaling workgroup performance with Switching and Fast Ethernet
Technology Review • 3
in the current corporate structure and continue to use
the legacy LAN structures installed up to ten to fifteen
years ago. Accompanying this emergence of these high
powered workgroups is the expanding presence of individual power users. Power users are high intensity networking individuals scattered throughout the corporate
enterprise. Together with the workgroups and power
users is a third phenomena in LAN architectures, the
clustering of heavy duty server systems in centralized
“farms”. These server farms are fast becoming the hub of
intense network activity.
Furthermore, these workgroups are often separated
from each other throughout a corporate “campus” and
require high speed connectivity. This need has led to the
emergence of high speed campus backbone structures.
Document Imaging Component Model
Input
Display &
Manipulation
Communication
Storage
Output
Standard Office Computing Platforms
Management Application
The Need for a Scalable
Migration Capability
These trends accentuated the need for highly flexible
and scalable LAN architectures, which are essential for
businesses today. It is neither necessary, nor cost effective,
to deploy the latest and fastest technology on every desktop and server. There are still many users, already connected to existing LANs, for whom shared 10Mbps bandwidth is quite sufficient. User needs are not growing uniformly. Different workgroups and individuals are expanding their needs at remarkably different rates.
4 • Scaling workgroup performance with Switching and Fast Ethernet
Conversely, there are network servers and high powered users for whom it is quite easy to justify greatly
expanded bandwidth up to 100Mbps. But this is not
always consistent. There may be workgroups for whom
10Mbps of dedicated bandwidth may be sufficient while
others may require 100Mbps. Scalability and migration,
therefore, refers to the ability to independently grow each
workgroup from shared legacy LANs operating at bandwidths of 10Mbps to LANs that can deliver bandwidths
of 100Mbps to individual users, workgroups or server
farms, while still maintaining current operations and
effectively utilizing current technologies.
Migration Requirements
To support the evolving need for flexible bandwidth
among a diverse user population, a migration architecture is needed that would enable the corporation to:
• Protect the existing investment in the installed base
by allowing a compatible mix of new and existing
technologies. This implies that the imbedded
cabling infrastructure, the existing hardware and
the existing network protocols must be maintained
in place and operate coherently with new technology upgrades.
• Upgrade users selectively by allowing a mix of low
bandwidth and high bandwidth users on the same
network.
• Build and mix LAN segments from 10Mbps to
100Mbps range.
• Allow the segmentation of LANs through the use
of switching hubs that allow individual users or
workgroups to have the specific bandwidth they
need.
• Upgrade to higher speed technologies, such as
ATM, as they emerge.
Technology Review • 5
Technology Options
Switching Hubs
Destination
Address = D
Technology Options for Increasing Performance in
the Workgroup
MAC
Header
Once the customer has determined the need to add
increased bandwidth, there are four viable options for
increasing workgroup performance:
• Switched Ethernet (segmented): A typical work
group today consists of a shared 10Mbps
repeater(s). By adding an Ethernet switch to the
network, and attaching the 10Mbps repeater(s) to
the switch ports, 10Mbps of bandwidth is made
available to each repeater (segment).
• Switched Ethernet (private): Each user is attached
to a switched 10Mbps port, providing dedicated
10Mbps bandwidth to each user.
• Shared Fast Ethernet: All users share 100Mbps in
the workgroup via a shared bus architecture.
• Switched Fast Ethernet: Dedicates 100Mbps of
bandwidth to each port. Switched Fast Ethernet should be used to relieve bottlenecks in the
building backbone, in a few small power workgroups or to provide backbone connections for
Super Servers. Switched 100Mbps to the desktop
is overkill for virtually all applications and will be
for the next several years. However, a small percentage of isolated workgroups and super servers
are able to utilize the full bandwidth today.
6 • Scaling workgroup performance with Switching and Fast Ethernet
A
B
C
D
Figure 2 - Switching Hubs
The operation of switching hubs is easy to understand conceptually. As shown in Figure 2, when a frame
comes in to a port, the switch assembles enough of the
frame to recognize the destination MAC address. Once
the switch knows the destination address, it forwards the
frame only to the destination port.
In switched Ethernet, the bandwidth is no longer
shared among the stations and each station has the full
10Mbps LAN bandwidth available at all times. Thus, if
there are 10 stations on a 10Mbps Ethernet switch, they
can generate in the aggregate, up to 100Mbps down a
high speed link. The more stations, the greater the aggregate bandwidth.
Switched Ethernet (segmented)
One of the basic techniques for increasing LAN
bandwidth, and therefore improving performance, is to
reduce the number of competing users on a LAN by
switched Ethernet segmentation, or separating the LAN
into multiple segments. To allow communications among
users on different LANs, these segments are interconnected with a switch. These interconnected LANs appear to
be a single logical LAN for purposes of communication
between the connected servers and end stations.
To make a distinction between the overall interconnected LAN and individual physical LANs, the individual
Technology Review • 7
physical LANs are referred to as LAN segments or, in
Ethernet, collision domains and this technique for
improving LAN performance is called LAN segmentation (Figure 3). The carrier signal is not propagated
across bridges/routers thus collisions are limited to individual segments.
To Backbone and
Server Farms
Ethernet
Switch
Ethernet
Hub
Ethernet
Hub
Local
Server
Ethernet
Hub
Figure 3 - Segmentation
In all of these arrangements, the LANs continue to
operate in what is essentially a Shared Ethernet mode.
That is, the LAN stations on any given segment share the
bus facility and compete with each other for service listening for competing carrier tone before transmitting.
The introduction of switches, however, opens the
door for a new approach to supporting end user devices.
This approach, called Switched Ethernet (private), is
based on the direct connection of users to ports on an
intelligent switch.
Switched Ethernet (private)
Switched Ethernet is an important phase in the evolution of workgroups. This approach employs intelligent
switches to provide direct LAN connectivity among
devices attached to the same switch. Using this approach,
each device has the sole use of its 10Mbps link to the
switch. This dedicated 10Mbps bandwidth, combined
with a high speed link to the server, dramatically
improves performance by removing server bottlenecks as
well as increasing network access for the client stations.
This has great value because the resultant increase in
8 • Scaling workgroup performance with Switching and Fast Ethernet
aggregate bandwidth is accomplished without having to
replace existing 10Mbps adapter cards or cabling.
Fast Ethernet
Fast Ethernet is designed as a direct and simple evolutionary extension of 10Base-T Ethernet. Fast Ethernet,
or 100Base-T, uses the same CSMA/CD protocol, and is
simply a scaled up version of 10Mbps Ethernet. It has the
same reliable, robust and economical technology already
in wide usage. Fast Ethernet is defined as IEEE 802.3u
and is also known as 100Base-T. Figure 4 shows the similarities between Ethernet and Fast Ethernet. Fast
Ethernet increases the peak bandwidth available to a station to 100Mbps and, if used in a switched mode, is
capable of supporting gigabit range aggregate bandwidth.
In this context, the older shared media LANs are now
sometimes referred to as the legacy LANs.
Applications
No Change
Applications
Management
No Change
Management
CSMA/CD MAC
No Change
CSMA/CD MAC
AUI Interface
MII Interface
Customer Choice
Customer Choice
Thick Coax
(100Bases)
Thin Coax
(10Base2)
Fiber
(10Base-F)
Two-Pair
Category 3, 4, 5 UTP
(10Base-T)
Ethernet Physical Layer (PHY)
Options
Fiber
(100Base-FX)
Four-Pair
Category 3, 4, 5 UTP
(100Base-T4)
Two-Pair
Category 5 UTP, STP
(100Base-TX)
Fast Ethernet Physical Layer (PHY)
Options
Figure 4 - Compatible Ethernet and
Fast Ethernet Architectures
Fast Ethernet provides a migration path from
10Mbps to 100Mbps for LAN segments that require the
additional speed or capacity. The 100Mbps addresses
Technology Review • 9
peak bandwidth issues. In a shared Ethernet environment,
going from 10Mbps to 100Mbps increases the capacity
tenfold. In addition, and perhaps more importantly,
100Mbps also increases the speed of transmission (i.e. the
peak rate) by tenfold. A complex image file of 20Mbytes
which may take 20 seconds or more to send on 10Base-T,
would take only 2 seconds with 100Base-T. As more
applications come on-line with huge files sizes and the
need for fast reception, 100Base-T becomes essential.
This distinction between an aggregate need, as opposed
to a peak need is one of the determinations a designer
should consider in choosing between switched Ethernet
and Fast Ethernet. Network design factors will be
discussed later.
Cabling System Alternatives
One of the key values of Fast Ethernet is its ability
to operate on the same media as existing Ethernet LANs.
Fast Ethernet has several categories:
• 100Base-T4 —Supports 4 pair category 3, 4 or 5
unshielded twisted pair (UTP)
• 100Base-TX—Supports data grade 2 pair shielded
twisted pair (STP) wire and 2 pair category 5 UTP
• 100Base-FX—Supports 2 strand optical fiber
The flexibility of these specifications allows 100BaseT to be implemented in virtually any 10Base-T cabling
environment. More importantly, 100Base-TX, T4 and
FX can be mixed together and interconnected through a
hub stack, just as 10Base2 and 10Base5 coaxial Ethernet
can interoperate with 10Base-T twisted pair Ethernet.
10 • Scaling workgroup performance with Switching and Fast Ethernet
Adapter automatically shifts from 10Mbps ➝100Mbps mode
10/100 Switch
100Mbps Hub
Figure 5 - 10/100Mbps Auto-Negotiation
10/100Mbps Auto-Negotiation
To provide easy migration from 10 to 100Mbps, the
100Base-T standard includes automatic speed sensing as
part of the Auto-Negotiation, or Nway, function (Figure
5). This optional function allows an adapter capable of
transmitting at both 10Base-T and 100Base-T speeds to
automatically communicate available modes of operation
and use the fastest one supported by the device at the
other end.
Auto-Negotiation can be used in dual-function
10/100Mbps Ethernet adapters. The process happens
out of hand, with no loss of network throughout. To
begin, a 100Base-T station advertises its capabilities by
sending a burst of link integrity test pulses called a fast
link pulse (FLP), generated automatically at power up.
If the receiving station is a hub capable of 10Base-T
communication only, the FLPs will be ignored and the
segment will operate as 10Base-T. But if the hub can
support 100Base-T operation, it will detect the FLPs, use
the Auto-negotiation algorithm and FLP data to determine the highest possible segment speed, and send FLPs
to the adapter to automatically place the adapter into
100Base-T mode.
The change occurs with no manual or software
intervention. (If necessary, a network manager can use
management software to force the segment to operate
at 10Mbps even if both devices are capable of 100BaseT communication.)
Technology Review • 11
Topology Rules and Repeaters
There are certain characteristics of Fast Ethernet
that users need to consider when planning migration from
existing legacy Ethernet LANs. Fast Ethernet preserves
the 100 meter maximum UTP cable length from the hub
to the desktop. But as a result of scaling the MAC interface, other 100Mbps topology rules are different.
As in Ethernet LANs, 100Base-T LANs will be segmented into multiple collision domains. Each port on a
bridge or switch begins another collision domain. The
diameter of each such domain will depend on the media
and type of hub.
The 100Base-T standard defines two classes of
repeater, called Class I and Class II. Any domain can
contain, at most, one Class I or two Class II repeaters.
• Class I is predominantly used between media using
different signaling system, e.g., between T4 and TX
and T4 and FX as well as between T4 devices and
only one such repeater is allowed in a collision
domain. The maximum diameter of the collision
domain is 261 meters in a mixed copper (UTP) and
fiber environment.
• Class II is generally used between media using the
same signaling, e.g., TX to TX, FX to FX and TX
to FX (they use the same signaling). Two repeaters
are allowed in this mode. The illustration shows
various Class I and Class II repeater configurations.
With two Class II repeaters, the maximum diameter of the collision domain is 205 meters using
copper (UTP) connections. With one repeater, the
diameter can be extended to 309 meters, using a
fiber downlink.
Additional topology rules include:
• MAC to MAC (such as switch to switch) connec-tions using half duplex 100Base-FX fiber can be
extended to 412 meters.
• A pre-standard full duplex version of 100Base-FX
can be extended up to two kilometers.
12 • Scaling workgroup performance with Switching and Fast Ethernet
The illustration below (Figure 6) shows these rules
and provides an example of how they support large scale
networks.
100Base-T
switch, bridge
100m UTP
MAC - MAC:
412m fiber
100Base-T
repeater
One repeater:
309m (typically
209m fiber+
100 UTP)
100m UTP
100Base-T
repeater
5 meter UTP
Two repeaters:
205m (typically
100+5+100)
100m UTP
100m UTP
100Base-T
repeater
Bridge,
router or
switch
2Km full-duplex fiber
Figure 6 - 100Base-T Topology Rules
Planning and Provisioning the
Network
Network administrators need to be proactive in determining their need for increased bandwidth. This involves
an assessment of the current utilization of the network as
well as planning for future growth and application
requirements. In evaluating new alternatives for future
LAN migration the most important information needed
by the designer includes:
• Are your users currently complaining about poor
response times?
• Are you currently monitoring the average utilization of the network? You should plan to use netTechnology Review • 13
work management tools to actively monitor network utilization levels.
• Price Sensitivity
• Manageability and control requirements
• What are your plans for network expansion?
• What applications are you currently running?
What new applications do you plan on adding
over the next 2-3 years?
• Where are your servers located? Do they generally
serve one workgroup, or are they expected to serve
users from multiple workgroups?
• What type of infrastructure is in place: computer
bus, bandwidth, cable plant, server location?
Since it is generally accepted as good practice to
plan for PC upgrades on a 2-3 year cycle to maintain
competitiveness, network administrators should use a 3
year window for planning the introduction of new applications and network infrastructure requirements. A predictive performance model should be developed to provide a “what if ” estimate of network utilization levels
and end-user response times.
Over the last ten years, Ethernet LANs have traditionally been built utilizing shared 10Mbps technology,
and to date this infrastructure has served the majority of
customers very well. The combination of high performance end systems, and low-cost high-performance networking solutions make the time right for network
administrators to re-evaluate their workgroup design
strategies.
Existing Infrastructure
The first step in determining whether to utilize
switched Ethernet or Fast Ethernet is to take a look at the
existing infrastructure, and planned expansion. A key
part of the evaluation is taking a look at the cabling type,
PCs and Servers deployed in the network. In many environments, the decision will be obvious, based on cabling
and how much bandwidth the PCs and Servers attached
to the LAN can accommodate. In other environments,
the decision will be less obvious and will need to be based
on other factors such as traffic patterns and network utilization, performance requirements, manageability and
control requirements, and price sensitivity.
One of the first things customers should consider
when looking to increase performance in the workgroup
is the type of cabling installed.
As shown in the following table, Fast Ethernet can
be run over Category 3 (4 pair), Category 4 (4 pair) or
Category 5 (2 or 4 pair) cable. If only 2 pairs are available, deploying Fast Ethernet is not an option, and
Switched Ethernet should be used.
Cabling Options
10Base-T 100Base-TX
# of Pairs required
Cable Category
100Base-T4 100Base-FX
2
2
4
Cat 3/4/5
Cat 5
Cat 3/4/5
Fiber
Choosing the Right Technology for the Workgroup
The decision as to which technology to use should be
based on the following criteria:
• Existing Infrastructure
• Current and planned applications
• Traffic patterns and network utilization in the
workgroup
14 • Scaling workgroup performance with Switching and Fast Ethernet
When looking at increasing performance in workgroups with either ISA or PCMCIA machines, Switched
Ethernet is the solution. The table on page 16 shows the
actual amount of bandwidth that can be supported by
various bus types. The actual amount of bandwidth
available to a NIC is impacted by other adapters such as
SCSI, IDE, and VGA that are attached to the bus. Since
the actual amount of bandwidth available to an ISA or
Technology Review • 15
PCMCIA NIC is approximately 11Mbps, it cannot take
advantage of 100Mbps of bandwidth. Switched Ethernet
is the solution to increasing performance in these environments. With Switched Ethernet, organizations receive
a significant performance boost without incurring the
tremendous cost of replacing the ISA or PCMCIA
machines deployed in the network. Although MCA can
support upwards of 80Mbps, no vendor has a MCA Fast
Ethernet card. Therefore, the solution for increasing performance to a MCA system is also Switched Ethernet.
ISA
Bus Bandwidth
(theoretical)
66Mbps
PCMCIA
66Mbps
EISA
264Mbps
MCA
PCI
320Mbps 1056Mbps
Bus Bandwidth 10-20Mbps 10-20Mbps
(actual)
64Mbps
Recommended
High Speed
Connection
Switched Switched Switched
Ethernet/ Ethernet Ethernet/
Fast
Fast
Ethernet
Ethernet
Switched Switched
Ethernet Ethernet
80Mbps
264Mbps
Current and Planned Applications
Before deciding which NIC to purchase for new PCI
machines, customers should review their planned roll-out
of new applications. As detailed in the table on page 17,
many applications that will be deployed over the next
several years can be supported by dedicated 10Mbps to
the desktop. If the environment does not require
100Mbps over the next 2-3 years, 10Mbps PCI adapters
should be installed. If 100Mbps will be required, then a
10/100 NIC should be installed. Whether these machines
should be installed using Switched Ethernet or Shared
Fast Ethernet depends on the applications, traffic
patterns and performance requirements in the workgroup. Many organizations are starting to deploy these
machines in 10Mbps mode with plans to expand to Fast
Ethernet in the future.
16 • Scaling workgroup performance with Switching and Fast Ethernet
Segmented
Ethernet
Switched
(dedicated)
Ethernet
Shared
Fast
Ethernet
E-mail
✓
✓
✓
Calendar
✓
✓
✓
Word-processing
✓
✓
✓
Business Graphics
✓
✓
✓
Video on demand
(for training purposes)
✓ *
✓
Video-conferencing
✓ **
✓
Image transfer
✓
✓
Distributed database
✓
✓
Graphics
✓
✓
Multimedia
✓
✓
High definition, 3D modeling
✓
Sophisticated CAD/CAM
✓
*(1.2Mbps required)
**(2.5Mbps required)
A key part of increasing the performance in the
workgroup is eliminating the bottlenecks to servers. As
with PCs, the solution for eliminating server bottlenecks
depends on the servers throughput capacity. Servers can
be classified into three basic categories: PC Desktop, PC
server, and Super Server. PC servers have traditionally
been standard PCs with the latest high performance
processor, a bit of extra RAM and maybe a larger hard
disk or two. Based on either ISA or EISA, these PCs used
as servers can often be accommodated by having a direct
10Mb/sec connection to a switch. These servers will likely be local to the workgroup, and expected to serve only
one workgroup. High performance servers, based on
EISA or PCI, with a Pentium processor, PCI bus, high
performance drive controller and multiple disk drives can
generally sustain throughput of 25-30Mbps. In these
cases dedicated 100Mbps cannot be fully realized, and
multiple servers can be connected with shared fast Ethernet, without restricting performance. However, super
servers, specially architected servers with PCI buses,
Technology Review • 17
1993
1994
1995
1996
1997
1998
PC Desktop Server
installed base
59
56
52
47
44
40
PC Server
% installed base
13
19
26
32
38
42
Super Server
% installed base
2
2
3
3
4
4
160
140
Workgroup Throughput
(Mbps)
multiple processors, and RAID-based disk subsystems,
warrant the highest performance connections in a network, either to a small shared Fast Ethernet backbone
segment that comprises a small server farm, or directly
to a Fast Ethernet switch port.
Shared 10,
1 Server
120
Switched 10,
100Mb Server
100
80
Switched 10,
2 100Mb Servers
60
40
Shared 100,
2 Servers
20
0
0
4
8
12
16
20
# Simultaneous Server Reads
24
Source: The
Tolly Group
Figure 7 - Switched Ethernet and
Shared Fast Ethernet Performance
Source: IDC LAN Server Market, 1995
The table above shows the installed base of various
server types. The majority of the installed base will not
require greater than shared 100Mbps today. A small percentage of servers (Super Servers) installed will benefit
from dedicated 100Mbps.
Traffic Patterns and Performance Requirements in
the Workshop
The graph on the next page (Figure 7) summarizes
the results of tests conducted by the Tolly Group, and
highlights some of the differences between switched
Ethernet and shared Fast Ethernet. As you can see in the
table above, both technologies provide dramatic performance improvements over traditional 10Mbps Ethernet,
but the exact performance level is a function of 1) the
number of servers in the workgroup and 2) the number
of clients trying to access the server at any given moment
in time.
18 • Scaling workgroup performance with Switching and Fast Ethernet
In a single-server environment, the shared Fast Ethernet network shows similar performance to the 10Mbps
switched network once several clients access the server
simultaneously. With a few clients, the Fast Ethernet solution will provide more throughput than a 10Mbps
switched environment because 8 or more clients accessing the server and running at 10Mbps are required to
even generate the traffic possible from one high performance 100Mbps node. Once more users start accessing
the network, the performance is the same, because the
network bottleneck has shifted from the clients to the
100Mbps server link.
If the performance problem on the LAN is due to
a few users occasionally transferring very large files that
require peak bandwidth (power workgroups, sophisticated CAD applications, 3D modeling, etc.) then shared
Fast Ethernet to the desktop is the answer. If the problem
is due to too many users accessing the server or applications requiring predictable throughput, then the decision
should be made based on control and security requirements.
In a dual server workgroup, (or local server and backbone server(s)) the switched network has the potential to
provide up to twice the throughput of the Fast Ethernet
network, if the switch supports two servers on separate
switched 100Mbps ports. Whether the switched network
Technology Review • 19
realizes the potential performance advantage is a function of the number of client PCs accessing the server at
any moment. When less than twelve users access the
servers at any moment, this configuration is similar in
performance to the single server solution. However with
very high aggregate traffic patterns in the workgroup, this
solution can provide dramatic performance improvements. If congestion is due to large workgroups with
multiple users placing a steady, uniform load on the network (client server applications such as transaction/database processing, document management, video based
training, video conferencing), then switched Ethernet
should be used. Figure 8 shows a decision tree to assist in
choosing the correct workgroup technology.
across multiple switches in a network.) Therefore, while in
some environments the performance of Fast Ethernet
may exceed switched Ethernet, network designers may
choose dedicated switched Ethernet purely because of the
benefits of virtual LANs.
Cost
ISA Bus
PCs
Office
Automation
Computer
Type
PCI,
EISA, Unix
workstation
Applications
Existing Network
Installed PCs
CAD,
Graphics
Network
Status
Manageability and Control Requirements
Improving network security has ranked as a major
concern and goal of network managers. In some cases
customers may be willing to trade some performance
gains for increased security. Unlike repeaters, which forward data to all ports, switches forward data only to the
port where the destination mode is located. Users on
other ports never see that data, making LAN traffic more
secure while increasing total network throughput. For
example, a finance department and a sales department
can be connected to different switch ports, and not be
able to see or be affected by each others data. Virtual
LANs further enhance these benefits.
Virtual LANs, supported in advanced switches like
3Com’s LinkSwitch 1000 & 3000, allow network administrators to enhance network performance, improve network security, and reduce the administrative burden of
adds, moves, & changes. The power of virtual LANs
grow as more switching is added to networks, with the
highest benefits accruing to networks providing dedicated
switched Ethernet ports to each user. (For example,
broadcasts in this environment could be controlled to
only those members of the specific virtual LAN’s designated to receive it, even though the users may be spread
20 • Scaling workgroup performance with Switching and Fast Ethernet
Performance
Manageability,
control,
security
Requirements
Performance
New Network
or new PCs
PCI, EISA,
Unix
workstation
Computer
Type
ISA Bus
PCs
Segmented
Ethernet
Requirements
CAD,
Graphics
Applications
Requirements
Private
Ethernet
2 Pair Cat 3
cabling
Infrastructure
Fast
Ethernet
Cat 5, or 4-pr Cat 3 cabling
Private
Ethernet
Manageability,
control,
security
Office
Requirements
Automation
Cost
Segmented
Ethernet
Applications
CAD,
Graphics
Upgrade
Computer
Figure 8 - Choosing the Right Workgroup Technology:
Decision Tree
The Bottom Line
Today’s enterprise networks are increasingly dependent on complex, high performance LANs that supply
sufficient bandwidth right to the desktop to support a
myriad of graphics and server applications. The remarkable synergy between Shared Ethernet, Switched Ethernet and Shared Fast Ethernet provides the basis for a
range of scalable solutions that meet a wide variety of
networking concerns. Whether the network has to supTechnology Review • 21
port the aggregate bandwidth requirements of an
increased population of users or provide the peak horsepower to meet the demands of high performance workgroups, these three complementary technologies provide
a comprehensive, scalable, integrated workgroup
solution. By preserving the wiring infrastructure, protocols and ethernet interfaces already in place, they offer a
cost effective, easy way to implement a migration plan for
the next decade.
As a network manager begins to consider these
issues, it is important to do some preliminary analysis
based on the decision tree illustrated in Figure 8. This
flow chart approach should provide guidance in zeroing
in on just the right solution for each separate workgroup.
One critical consideration, as this migration strategy
takes shape, is to remember that the best way to develop
a comprehensive and assured solution is to identify a supplier that can furnish a full range of product capabilities
to meet all of the possible needs the company may have.
Combined with full design support, as well as provisioning advice and assistance, that supplier can help managers meet their performance goals at the least possible
cost and with the greatest success.
Glossary
10Base-T — The IEEE 802.3 specification for ethernet
over unshielded twisted pair (UTP).
100Base-FX — 100Mbps Ethernet implementation
over fiber.
100Base-T4 — 100Mbps Ethernet implementation
using 4 pair Category 3, 4 or 5 cabling.
100Base-TX — 100Mbps Ethernet implementations
over Category 5 and Type 1 cabling.
Adapter — A board installed in a computer system to
provide network communication capabilities to and from
that computer system. Also called a Network Interface
Card (NIC).
Adapter Card — Circuit board or other hardware that
provides the physical interface to the communications
network.
Asynchronous Transfer Mode (ATM) — The
CCITT standard for cell relay wherein information for
multiple types of services (voice, video, data) is conveyed
in small, fixed-size cells. ATM is a connection oriented
technology used in both LAN and WAN environments.
ATM Downlink — A 155Mbps vertical link containing
one or more ATM virtual channels.
Attachment Unit Interface (AUI) — An IEEE 802.3
cable connecting the MAU (media attachment unit) to
the network device. The term AUI also can be used to
refer to the host back-panel connector to which an AUI
transceiver cable might attach.
Backbone — The part of a network used as the primary path for transporting traffic between network segments.
Backplane — The main bus that carries data within a
device.
Bandwidth — Measure of the information capacity of
a transmission channel.
22 • Scaling workgroup performance with Switching and Fast Ethernet
Glossary • 23
Bridge — A device that connects and passes packets
between two network segments. Bridges operate at Layer
2 of the OSI reference model (the data-link layer) and
are insensitive to upper-layer protocols. A bridge will
examine all frames arriving on its ports and will filter,
forward or flood a frame depending on the frame’s Layer
2 destination address.
Bridge/Router — A device that can provide the functions of a bridge, router or both concurrently. Bridge/
router can route one or more protocols, such as TCP/IP
and/or XNS, and bridge all other traffic.
Bridging — Techniques for interconnecting two LAN
segments that utilize the same LLC procedures but may
use the same or different MAC procedures.
Carrier Sense Multiple Access/Collision
Detection (CSMA/CD) — A channel access mechanism wherein devices wishing to transmit first check the
channel for a carrier. If no carrier is sensed for some
period of time, devices can transmit. If two devices
transmit simultaneously, a collision occurs and is detected
by all colliding devices, which subsequently delays
their retransmissions for some random length of time.
CSMA/CD access is used by Ethernet and IEEE 802.3.
Category 3 Unshielded Twisted Pair (CAT-3) —
Industry standard for unshielded twisted wire pair capable of supporting voice and low-grade data traffic.
Category 5 Unshielded Twisted Pair (CAT-5) —
Industry standard for unshielded twisted wire pair capable of supporting high speed data traffic over short (LAN
and CAN) distances.
Collapsed Backbone — A non distributed backbone
where all network segments are interconnected via an
internetworking device. A collapsed backbone may be
a virtual network segment existing in a device such as a
hub, a router or a switch.
Concentrator — Device that serves as a wiring hub in
star-topology network. Sometimes refers to a device containing multiple modules of network equipment.
24 • Scaling workgroup performance with Switching and Fast Ethernet
Congestion — Excessive network traffic.
Congestion Control — Network management issue
for the controlling of traffic flow so switches and endstations are not overwhelmed with information and cells
subsequently dropped.
Contention — Network access method where devices
compete for the right to access the physical medium.
Cut-Through Switching — Refers to a method of
Frame Switching where the switching device commences
forwarding a frame after it has determined the destination port without waiting for the entire frame to have
been received on the incoming port. Also known as
on-the-fly switching.
Data Link Layer — Layer 2 of the OSI reference
model. This layer takes a raw transmission facility and
transforms it into a channel that appears, to the network
layer, to be free of transmission errors. Its main services
are addressing, error detection and flow control.
Dedicated LAN — Network segment allocated to a
single device. Used in LAN switched network topologies.
Electronic Data Interchange (EDI) — Method for
passing orders, invoices and other transactions electronically between locations or organizations.
End System — End-user device on a network. Also, a
nonrouting host or node in an OSI network.
Enterprise Network — A geographically dispersed
network under the auspices of one organization.
Ethernet — A baseband LAN specification invented
by Xerox Corporation and developed jointly by Xerox,
Intel, and Digital Equipment Corporation. Ethernet networks operate at 10Mbps using CSMA/CD to run over
coaxial cable. Ethernet is similar to a series of standards
produced by IEEE referred to as IEEE 802.3.
802 — A set of IEEE specifications for local area networks (LANs) and metropolitan area networks (MANs).
802.1: general management and internetwork operations
such as bridging. 802.2: sets standards at the logical link
control sublayer of the data link layer. 802.3 —
Glossary • 25
CSMA/CD (Ethernet) standards, which apply at the
physical layer and the media access control (MAC) sublayer. 802.4: token passing bus standards. 802.5: token
ring standards. 802.6: MAN standards. IEEE 802 standards become ANSI standards and are usually accepted
as international standards.
Frame — A logical grouping of information sent as a
link-layer unit over a transmission medium. The terms
packet, datagram, segment, and message are also used
to describe logical information groupings at various
layers of the OSI reference model and in various technology circles.
Hardware Address — Also called physical address or
MAC-layer address, a data-link layer address associated
with a particular network device. Contrasts with network
or protocol address which is a network layer address.
Hub — Common name for a repeater. Strictly, it is a
non-retiming device.
IEEE 802.3 — IEEE LAN protocol that specifies an
implementation of the physical layer and MAC sub layer
of the link layer. IEEE 802.3 uses CSMA/CD access at a
variety of speeds over a variety of physical media. One
physical variation of IEEE 802.3 (10Base5) is very similar
to Ethernet.
Institute of Electrical and Electronic Engineers
(IEEE) — Professional organization that defines network
standards. IEEE LAN standards are the predominant
LAN standards today, including protocols similar or
virtually equivalent to Ethernet and Token Ring.
Internet Address — Also called an IP address. It is
a 32-bit address assigned to hosts using TCP/IP. The
address is written as four octets separated with periods
(dotted decimal format) that are made up of a network
section, an optional subnet section, and a host section.
Internetwork — A collection of networks interconnected by routers that functions (generally) as a single
network. Sometimes called an internet, which is not to
be confused with the Internet.
26 • Scaling workgroup performance with Switching and Fast Ethernet
Internetworking — General term used to refer to the
industry that has arisen around the problem of connecting networks together The term can refer to products,
procedures, and technologies.
Kilobits per second (Kbps) — Thousand bits per
second. A measure of transmission speed.
Latency — The delay between the time a device
receives a frame and the frame is forwarded out of the
destination port.
Local Area Network (LAN) — A network covering a
relatively small geographic area (usually not larger than a
floor or small building). Compared to WANs, LANs are
usually characterized by relatively high data rates.
LAN Segmentation — Dividing LAN bandwidth into
multiple independent LANs to improve performance.
Management Information Base (MIB) — A database of information on managed objects that can be
accessed via network management protocols such as
SNMP and CMIP.
Media Access Control (MAC) — A method of controlling access to a transmission medium. For example,
token ring, Ethernet, FDDI, etc.
Media Access Control Layer Address (MAC
Layer Address) — Also called hardware address or
physical address. A data-link layer address associated with
a particular network device. Contrasts with network or
protocol address which is a network layer address.
Media Access Control Sub Layer (MAC Sub
layer) — As defined by the IEEE, the lower portion of
the OSI reference model data link layer. The MAC sub
layer is concerned with media access issues, such as
whether token passing or contention will be used.
Microsegmentation — Division of a network into
smaller segments usually with the intention of increasing
aggregate bandwidth to devices.
Network Interface Card (NIC) — The circuit board
or other hardware that provides the interface between a
communicating DTE and the network.
Glossary • 27
Network Layer — Layer 3 of the OSI reference model.
Layer 3 is the layer at which routing occurs.
Network Management System (NMS) — A system
responsible for managing at least part of a network.
NMSs communicate with agents to help keep track of
network statistics and resources.
Packet — A logical grouping of information that
includes a header and (usually) user data.
Packet Buffer — Storage area to hold incoming data
until the receiving device can process the data.
Protocol — A formal description of a set of rules and
conventions that govern how devices on a network
exchange information.
Protocol Address — Also called a network address. A
network layer address referring to a logical, rather than a
physical, network device.
Protocol Stack — Related layers of protocol software
that function together to implement a particular communications architecture. Examples include AppleTalk and
DECnet.
Repeater — A device that regenerates and propagates
electrical signals between two network segments.
Router — An OSI Layer 3 device that can decide which
of several paths network traffic will follow based on some
optimality metric. Also called a gateway (although this
definition of gateway is becoming increasingly outdated),
routers forward packets from one network to another,
based on network-layer information.
Routing — The process of finding a path to the destination host. Routing is very complex in large networks
because of the many potential intermediate destinations
a packet might traverse before reaching its destination
host.
Routing Bridge — MAC-layer bridge that uses network layer methods to determine a network’s topology.
28 • Scaling workgroup performance with Switching and Fast Ethernet
Simple Network Management Protocol (SNMP)
— The Internet network management protocol. SNMP
provides a means to monitor and set network configuration and runtime parameters.
Store and Forward — Switching technique where
frames, packets or messages are temporarily received and
buffered at intermediate points between the source and
destination. Bridges, Routers, X.25 switches and ATM
switches are all based on store and forward technology.
Subnetwork — Collection of OSI end systems and
intermediate systems under the control of one administrative domain and using a single network access protocol. For example, private X.25 networks, a series of
bridged LANs.
Switch — In the context of Frame or LAN switching,
this refers to a device which filters, forwards and floods
frames based on the frame’s destination address. The
switch learns the addresses associated with each switch
port and builds tables based on this information to be
used for the switching decision. Some switches are high
speed implementations of bridges where switching decisions are made in silicon, usually an Application Specific
Integrated Circuit (ASIC).
Switching Hubs — Hubs that use intelligent Ethernet
switching technology which interconnects multiple
Ethernet LANs and higher speed LANs, such as FDDI.
Transmission Control Protocol/Internet Protocol
(TCP/IP) — The common name for the suite of protocols developed by the U.S. Department of Defense in the
1970s to support the construction of worldwide internetworks. TCP and IP are the two best-known protocols in
the suite. TCP corresponds to Layer 4 (the transport
layer) of the OSI reference model. It provides reliable
transmission of data. IP corresponds to layer 3 (the network layer) of the OSI reference model and provides
connectionless datagram service.
Glossary • 29
Twisted Pair (TP) — Cable consisting of two 18 to 24
AWG (American Wire Gauge) solid copper strands twisted around each other. The twisting provides a measure of
protection from electromagnetic and radio-frequency
interference.
NOTES
Unshielded Twisted Pair (UTP) — Four-pair wire
medium used in the transmission of many different protocols such a Ethernet, 10BaseT, and CDDI.
Wide Area Network (WAN) — A network which
encompasses interconnectivity between devices over a
wide geographic area. Such networks would require
public rights-of-way and operate over long distances.
Workgroup Switching — The ability to handle a
symmetric traffic patterns via high-speed (100Mbps)
interface and intelligent switching.
30 • Scaling workgroup performance with Switching and Fast Ethernet
Notes • 31
This Technology Guide is one
of a series of guides, published
by ATG, designed to put complex
communications and networking
technology concepts into practical
and understandable terms.
Each guide provides objective,
non-biased information to assist in
Scaling Workgroup
Performance with
Switching and
Fast Ethernet
the internal education, evaluation
and decision making process.
This Technology Guide, as well
as the other Communications and
Networking Technology Guides
in the series, are available
on ATG‘s Web Site.
http://www.techguide.com
Produced and Published by
One Apple Hill, Suite 216, Natick, MA 01760
Tel: (508) 651-1155 Fax: (508) 651-1171 E-mail: [email protected]
ATG’s Communications &
Networking Technology
Guide Series
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement