Network Hardware Concepts - McGraw-Hill

Network Hardware Concepts - McGraw-Hill
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 1
CHAPTER
Network Hardware
Concepts
In this chapter you will learn about
• The concepts of gateways, routers, and bridges
• The role of client/server computing
• Firewalls and proxy servers
• Database servers
• Application servers
• Mail servers
• FTP servers
• File and print servers
• Fax servers
• Web servers
• How failover, clustering, scalability, and high availability relate to a
network server
• The role of network interface cards (NICs) and the concepts of
Adaptive Fault Tolerance, Adapter Load Balancing, and Adapter Teaming
• The characteristics of Ethernet, Fast Gigabit Ethernet, and Token Ring
networks
While individual computers can be quite powerful, they are still “individual.” Sharing
files and resources typically meant copying a file to a diskette, then manually walking
that diskette to other systems—for example, working on a document after work, then
returning that updated document to work the next day in order to print it. Obviously,
this is a cumbersome and time-consuming process. If there were a means of “connecting” two or more computers, you could access your work from another location
1
1
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 2
Server+ All-in-One Certification Exam Guide
2
(i.e., a computer in your home), finish the work that night, and then send the work to
a printer located back at the office. This is the underlying premise behind a network—
two or more computers connected together in order to share files, resources, and even
applications. This chapter introduces you to the basic concepts and terminology
needed to understand the tangible elements of common networks and servers.
A Network Primer
A networked computer that provides resources is called a server. The computer accessing those resources is referred to as a workstation or client. Servers are usually the most
powerful computers on the network because they require the added processing power
to service the many requests of other computers sharing their resources. By comparison, workstations or clients are usually PCs that are cheaper and less powerful. As a
rule, a computer may be a server or a workstation, but rarely both (this separation
greatly simplifies the management and administration of the network). Of course, all
the computers on a network must be physically connected, and such connections are
typically established with network interface card (NIC) adapters and copper (or fiberoptic) cabling.
Advantages of a Network
With individual computers, applications and resources (such as printers or scanners) must be duplicated between PCs. For example, if two data analysts want to
work on an Excel spreadsheet and print their results each day, both computers will
need a copy of Excel and a printer. If the users needed to share data, it would have
to be shuttled between the PCs on diskette or CD-RW. And if users needed to share
computers, they would have to wade through the other user’s system—each with its
own desktop setup, applications, folder arrangement, and so on. In short, it would
be a wasteful, frustrating, and error-prone process. As more users become involved,
it wouldn’t take long before the whole process would be impossible to handle.
However, if those two computers in our example were networked together, both
users could use Excel across the network, access the same raw data, and then output
their results to a single “common” printer attached to the network. If more users
were then added to the network, all users could share the application, data, and
resources in a uniform fashion. More specifically, computers that are part of a network can share:
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 3
Chapter 1:
Network Hardware Concepts
3
• Documents (memos, spreadsheets, invoices, and so on)
• E-mail messages
• Word-processing software
• Project-tracking software
• Illustrations, photographs, videos, and audio files
• Live audio and video broadcasts
• Printers
• Fax machines
• Modems
• CD-ROM drives and other removable media drives (such as Zip and Jaz drives)
• Hard drives
Because many computers can operate on one network, the entire network can be efficiently managed from a central point (a network administrator). Consider the previous
example and suppose that a new version of Excel became available to our data analysts.
With individual computers, each system would have to be upgraded and checked separately. That’s not such a big deal with only two systems, but when there are dozens (even
hundreds) of PCs in the company, individual upgrades can quickly become costly and
inefficient. With a network, an application only needs to be updated on its server once—
then all the network’s workstations can use the updated software immediately. Centralized
administration also allows security and system monitoring to take place from one location.
Network Sizes
Computer networks typically fit into one of three groups depending on their size and
function. A local area network (LAN) is the basic classification of any computer network.
LAN architecture can range from simple (two computers connected by a cable) to complex (hundreds of connected computers and peripherals throughout a major corporation). The distinguishing feature of a LAN is that it is confined to a limited geographic
area such as a single building or department. If the computers are connected over several buildings across a large metropolitan area, the network is sometimes termed a metropolitan area network (MAN). By comparison, a wide area network (WAN) has no
geographical limit. It can connect computers and peripheral devices on opposite sides
of the world. In most cases, a WAN is made up of a number of interconnected LANs—
perhaps the ultimate WAN is the Internet.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 4
Server+ All-in-One Certification Exam Guide
4
Network Types
Networks are generally divided into two distinct categories: peer-to-peer and server based.
This is an important distinction because these two categories are vastly different and
offer different capabilities to the users. Peer-to-peer networks are simpler and less
expensive networks that appear in small organizations (such as SOHO or small workgroup applications). Server-based networks are found in mid-sized and larger organizations where security, centralized administration, and high traffic capacity are
important. Let’s look a bit closer at these network types.
Peer-to-Peer Networks
This is a simple and straightforward networking approach that simply connects computers to allow basic file sharing. There are no dedicated servers, and there is no hierarchy among the computers. Since all the computers are equal, they are known as peers.
Each computer serves as both a client and a server, and no administrator is responsible
for the entire network—the user at each computer determines what data on that computer is shared on the network. All users can share any of their resources in any manner
they choose. This includes data in shared directories, printers, fax cards, and so on.
Peer-to-peer networks are also commonly called workgroups (which implies a small
group of people) because there are typically 10 or fewer computers in a peer-to-peer
network. As a result of this simplicity, peer-to-peer networks are often less expensive
than server-based networks.
In a peer-to-peer network, the networking software does not require the same standard of performance or security as the networking software designed for dedicated
server systems. In fact, peer-to-peer networking capability is built into many popular
operating systems (such as Windows 98/Me). This means you can set up a peer-to-peer
network without any additional network operating system.
Security is a real weakness in peer-to-peer environments. Generally speaking, security
(i.e., making computers—and the data stored on them—safe from harm or unauthorized
access) consists of setting a password on a resource (i.e., a directory) that is shared on the
network. All peer-to-peer network users set their own security, and shared resources can
exist on any computer, so centralized control is very difficult to maintain. This has a big
impact on network security because some users may not implement any security measures at all. In summary, a peer-to-peer network is often the best choice when
• There are 10 (or fewer) users.
• Users share resources (such as files and printers) but no specialized servers exist.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 5
Chapter 1:
Network Hardware Concepts
5
• Security is not an issue.
• The organization (and the network) is expected to experience only limited growth.
NOTE: Because every computer in a peer-to-peer environment can act as both
a server and a client, users generally need additional training before they can
act as both users and administrators of their computers.
Server-Based Networks
In most network situations, the duality of peer-to-peer networks is simply not adequate.
Limited traffic capability and security/management issues often mean that networks
need to use dedicated servers. A dedicated server is a computer that functions only as a
server to provide files and manage resources; it is not used as a client or workstation.
Servers are optimized to handle requests from numerous network clients quickly and
ensure the security of files and directories. Consequently, server-based networks have
become the standard models for modern business networking. Server-based networks
are also known as client/server networks (sometimes denoted as two-tier architectures).
NOTE: Servers provide specific resources and services to the network, and
there may be several (perhaps many) servers available in a given network.
Server Types
As networks increase in size (i.e., as the number of connected computers increases, and
the physical distance and traffic between them grows), more than one server is usually
needed. Spreading the networking tasks among several servers ensures that each task
will be performed as efficiently as possible. Servers must perform varied and complex
tasks, and servers for large networks have become specialized to accommodate the
expanding needs of users. Some examples of different server types included on many
large networks are listed here:
• File and print servers File and print servers manage the user’s overall access and
use of file and printer resources. For example, when you’re running a wordprocessing application (such as Microsoft Word), that application runs on your
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 6
Server+ All-in-One Certification Exam Guide
6
workstation. The document stored on the file and print server is loaded into your
workstation’s memory so that you can edit or use it locally. In other words, file and
print servers are used for file and data storage. If you wish to print the document,
the file and print server manages the transfer of that file to the network printer.
• Database servers In most cases, a database server is a server that runs an SQLbased database management system (DBMS). Client computers send the SQL
requests to the database server. The server accesses the stored database to process
the request, and then returns the results to the client computer. When referring to
a database server, the term “server” may refer to the computer itself or the DBMS
software that manages the database (such as Microsoft SQL Server).
• Application servers Where file and print servers will download a file to the
requesting client PC, an application server does not—only the results of a request
are sent to the client PC. For example, you might search the employee database for
all employees who were born in November. Instead of the entire database being
downloaded to your PC so that you can search it, the search is performed on the
application server itself, and only the result of your query is sent from the server to
your computer. This subtle but powerful difference makes application servers
(such as Lotus Domino) ideal for maintaining vast quantities of data and efficiently providing that data to clients.
• Mail servers E-mail is an important part of modern communication, so mail
servers (such as Microsoft Exchange Server) handle the flow of e-mail and messaging between network users. In most cases, mail servers are similar to application
servers because the e-mail typically remains on that server. When you check your email, you only see the e-mail intended for your screen name. Storing e-mail in a
central fashion such as this allows for better security and e-mail management (i.e.,
old e-mails can be purged after so many days in a system-wide fashion).
A variation of this is the mailing list server (a.k.a., list server), which is needed for
creating, managing, and serving mailing lists. Stand-alone list servers (such as
Majordomo) generally offer more features and better performance than their integrated counterparts. Uses for mailing lists and list servers include the distribution
of e-zines, newsletters, product updates, technical support documents, classroom
schedules, and product brochures, along with discussion forums for clubs and
groups, electronic memos, and so on.
• Fax and communication servers Networks rarely exist in a vacuum, and there
are generally several ways to access the network from outside. Two popular means
of external network access are faxes and dial-up. A fax server (such as FaxMaker)
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 7
Chapter 1:
Network Hardware Concepts
7
manages fax traffic into and out of the network using one or more fax/modem
cards. This allows network users to send faxes outside of the network (and vice
versa). Communication servers handle data file and e-mail transfers between your
own networks and other networks, mainframe computers, or remote users who
dial in to the servers over modems and telephone lines. For example, a network
user may access the Internet through a communication server.
• Audio/video servers Audio and video servers deliver multimedia capabilities to
Web sites by giving users the ability to listen to sound or music and watch movie
clips through Web browser plug-ins. While the use of traditional formats like
.WAV, .MIDI, .MOV, or .AVI on Web sites doesn’t really demand a specialized
server, the recent emergence of streaming audio and video content has made the
audio/video server a necessity in many cases (with tools such as RealServer Plus).
New streaming technologies mark an important transition for multimedia on the
Web, and will undoubtedly become one of the Internet’s most exciting technologies as it evolves.
• Chat servers It is common practice for two or more users to exchange real-time
messages. This is called a chat, and chat servers (using tools like MeetingPoint) provide the management for real-time discussion capabilities for a large number of
users. Potential chat uses include teleconferences, private meeting areas, help support forums, and employee recreational get-togethers. The three major types of
communications servers are Internet Relay Chat (IRC), conferencing, and community servers. The most advanced chat servers have recently started augmenting the
text-based medium of conversation with dynamic voice (and even video) support.
It is common for IRC-based chat to use dedicated IRC servers (with software like
IRCPlus).
• FTP servers From downloading the newest software to transferring corporate
documents, a significant percentage of Internet traffic consists of file transfers. File
Transfer Protocol (FTP) servers make it possible to move one or more files between
computers with security and data integrity controls appropriate for the Internet
(using tools like ZBServer Pro). FTP is a typical client/server arrangement. The FTP
server does the main work of file security, file organization, and transfer control.
The client (sometimes part of a browser and sometimes a specialized program
such as FTP Voyager) receives the files and places them onto the local hard disk.
• News servers News servers function as a distribution and delivery source for over
20,000 public newsgroups currently accessible over the USENET news network
(the largest news and discussion group-based network on the Internet). News
servers use tools (like INN News Server) that employ the Network News Transport
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 8
Server+ All-in-One Certification Exam Guide
8
Protocol (NNTP) to interface with other USENET news servers and distribute news
to anyone using a standard NNTP newsreader (such as Agent or Outlook Express).
News servers also make it possible to serve your own news and discussion groups
publicly over the Internet—or privately over your own local network.
• Gateway servers A gateway is a translator that allows differing networks to communicate. For example, one common use for gateways is to act as translators
between personal computers and minicomputer or mainframe systems. In a LAN
environment, one computer is usually designated as the gateway computer. Special
application programs in the desktop computers access the mainframe by communicating with the mainframe environment through the gateway computer, and
users can access resources on the mainframe just as if those resources were on their
own desktop computers.
• Firewalls and proxy servers Simply stated, a firewall is a feature designed to prevent unauthorized access to or from a private network (i.e., a corporation’s LAN),
and is generally considered to be a first line of defense in protecting private information. Firewalls can be implemented in both hardware and software (and often
involve a combination of both). When properly implemented, firewalls prevent
unauthorized Internet users from accessing private networks that are connected to
the Internet—especially intranets. All messages entering or leaving the intranet
pass through the firewall, which examines each message and blocks those that do
not meet the required security criteria. There are numerous firewall techniques
including packet filters, application gateways, circuit level gateways, and proxy
servers. The proxy server is perhaps the most popular form of firewall. In actual
practice, a proxy server sits between a client program (i.e., a Web browser) and
some external server (usually another server on the Web). The proxy server effectively hides the true network address, then monitors and intercepts any requests
being sent to the external server, or that come in from the Internet connection.
This allows the proxy server to filter messages, improve performance, and share
connections.
Filtering is a security feature. Proxy servers can inspect all traffic (in and out) over
an Internet connection and determine if there is anything that should be denied
transmission, reception, or access. Since this filtering works both ways, a proxy
server can also be used to keep users out of particular Web sites by monitoring for
specific URLs, or restrict unauthorized access to the internal network by authenticating users. Since proxy servers are handling all communications, they can log
everything the user does. For HTTP (Web) proxies, this includes logging every URL.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 9
Chapter 1:
Network Hardware Concepts
9
For FTP proxies this includes tracking every downloaded file. A proxy server can
also examine the content of transmissions for “inappropriate” words or scan for
viruses. Proxy servers can also improve performance by proxy server caching. The
proxy server analyzes user requests and determines which (if any) should have the
content stored temporarily for immediate access. One example might be a company’s home page located on a remote server. If many employees visit this page
several times a day, the proxy server can cache it for immediate delivery to the Web
browser. Some proxy servers—particularly those targeted at small business—
provide a means for sharing a single Internet connection among a number of
workstations. While this has practical limits in performance, it can still be a very
effective and inexpensive way to provide Internet services (such as e-mail) to an
entire office.
• Web servers Web servers allow you to provide content over the Internet using the
Hypertext Markup Language (HTML). A Web server (with software like Microsoft
PWS) accepts requests from browsers like Netscape and Internet Explorer, and then
returns the appropriate HTML document(s) to the requesting computer. A number
of server technologies can be used to increase the power of the server beyond its
ability to simply deliver standard HTML pages—these include CGI scripts, SSL
security, and Active Server Pages (ASPs).
• Telnet/WAIS servers Telnet servers give users the ability to log on to a host computer and perform tasks as if they’re actually working on the remote computer
itself. Users can access the host system through the telnet server from anywhere in
the world using a telnet client application. Before the arrival of the Web, Wide Area
Information Server (WAIS) servers were critical for allowing users to perform
searches for keywords in files. While telnet and WAIS are really not that popular
today, network developers looking to broaden their selection of Internet services
may consider supporting telnet or WAIS services.
Server Software
One major issue that separates servers from peer computers is the use of software. No
matter how powerful a server may be, it requires an operating system (i.e., Windows
NT/2000 or Novell NetWare) that can take advantage of the server’s resources. Servers
also require their specific server applications in order to provide their services to the
network. For example, a Web server may use Windows NT and Microsoft PWS. It’s not
important for you to fully understand software issues at this point. Chapter 2 covers
network protocols and operating systems in more detail.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 10
Server+ All-in-One Certification Exam Guide
10
Client/Server Advantages There is little doubt that server-based networks are more
complicated to install and configure, but there are some compelling advantages over
peer-to-peer networks:
• Sharing Servers allow for better resource organization and sharing. A server is
intended to provide access to many files and printers while maintaining performance and security for the user. A server’s data and resources can be centrally
administered and controlled. This centralized approach makes it easier to find files
and support resources than would otherwise be possible on individual computers.
• Security In a server-based environment, one administrator can manage network
security by setting network policies and applying them to every user.
• Backups Backup routines are also simplified because only servers need to be
backed up (client/workstation PCs do not). Server backups can be scheduled to
occur automatically (according to a predetermined schedule) even if the servers are
located on different parts of the physical network.
• Fault tolerance Because data is mainly held on servers, fault-tolerant data storage (i.e., RAID) can be added to the servers to prevent data loss due to drive failures or system crashes. This creates a more reliable server subject to less downtime.
• Users A server-based network can support thousands of users. Such a large network would be impossible to manage as a peer-to-peer network, but current monitoring and network-management utilities make it possible to operate a
server-based network for large numbers of users.
Server Reliability Reliability is basically the notion of dependable and consistent
operation—the probability that a component or system will perform a task for a specified period of time. This includes the server as well as the network, and is often measured as a function of the time between system failures using the term mean time
between failure (MTBF). Data integrity and the capability to warn of impending hardware failures before they happen are two other aspects of reliability. Servers frequently
include reliability features such as redundant power supplies and fans, predictive failure analysis for hard drives (SMART), and redundant array of independent disks
(RAID) systems to ensure that a server continues to function and protect its data even
when trouble occurs. Other reliability features include the memory self-test at boot
time where the system detects and isolates bad memory blocks, as well as error checking and correcting (ECC) memory to improve data integrity.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 11
Chapter 1:
Network Hardware Concepts
11
NOTE: Reliability is a critical server issue and is absolutely vital to long-term
network operation. Large networks typically strive for 99.999 percent reliability
or better.
Server High Availability
A server must constantly be “up” and ready for immediate
use, allowing a user to access the resources they need in real time. This is the issue of high
availability. Another aspect of highly available servers is the capability to quickly recover
from a system failure (i.e., use a “hot spare” RAID disk to recover data from a failed drive).
Highly available systems may or may not use redundant components (such as redundant
power supplies), but they should support the hot swapping of key components. Hot
swapping is the ability to pull out a failed component and plug in a new one while the
power is still on and the system is operating. A highly available system has the capability
to detect a potential failure and transparently redirect or failover the questionable
processes to other devices or subsystems. For example, some SCSI drives can automatically move data from marginal sectors (i.e., sectors that produce occasional read errors)
to spare sectors without the operating system or the user being aware of the change.
In general, availability is measured as the percentage of time that a system is functioning and usable. For instance, a system that provides 99-percent availability on a 24
hours/day, 7 days/week basis would actually experience the loss of 88 processing hours
a year (unacceptable to many users). However, a 99.999 percent level of availability
translates to about 5.25 minutes of unscheduled downtime per year (though this level
of availability may be quite costly to achieve).
Server Scalability Computer customers of the past often bought mainframes twice
the size they needed in anticipation of future growth, knowing that they would eventually “grow into” the machine. Today it’s possible to select computers to fit the task now,
then add more equipment as needs demand—this is known as scalability. A scalable PC
has the capability to grow in size (capacity) and speed. Some machines offer limited
scalability by design, while some can grow to virtually any size needed. Scalability
includes the ability to add memory (RAM), add additional processors (i.e., for multiprocessing platforms), add storage (hard drives), and still work within the limitations
of the network operating system.
There is a subtle difference between upgrading and scaling. An upgrade is the replacement of an existing component with a faster or better component. Scaling a PC is the
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 12
Server+ All-in-One Certification Exam Guide
12
addition of more components for added capacity. For example, an ordinary PC may use
a single processor. It may be possible to upgrade that processor to a faster model, but it
is not possible to scale up the processing capacity with more processors—you’d need a
server with a multiprocessing motherboard for that. By contrast, virtually all PCs can
scale memory (RAM) simply by adding more DIMMs or RIMMs to the system. The
same concept holds true for your disk space—you can upgrade to a larger or faster disk,
or you could scale up the drive capacity by adding additional hard drives.
SMP and Parallel Processing Since processors are a key element of server performance and scalability, it is a good time to cover multiprocessing in a little more detail.
A symmetric multiprocessing (SMP) machine is a computer that utilizes two or more
processors. Each processor shares memory and uses only one copy of the operating system. SMP machines can scale by starting small (with only two processors), then adding
more processors as business needs and applications grow. Beyond CPUs, such computers typically have the ability to scale memory, cache, and disks. Currently, SMP
machines are designed to scale from 2-32 processors.
There are limiting factors to consider when dealing with SMP systems. While it may
seem possible to scale far more than 32 processors, that is often not the case. If you were
to start with two processors and add two more, a near 100-percent improvement may
result, but because there is only one operating system and all memory is shared, a diminishing return on performance will be realized as more processors are added. Most SMP
systems will show worthwhile improvements until they scale above eight processors (the
diminishing return also varies based on the operating system and the applications in use).
While UNIX systems with 16 or more processors are not uncommon today, Windows NT
scalability is commonly thought to be limited to about four CPUs. In addition, many
operating systems or database applications can only utilize the first 2GB of memory.
By comparison, some of the largest and most scalable systems in the world utilize
parallel processing technology. Parallel processing takes SMP a step further by combining
multiple SMP nodes. These nodes can work in parallel on a single application—usually
a database that is fully parallel-capable. Because each node has its own copy of the
operating system, and the nodes communicate through a specialized interconnection
scheme, adding additional nodes does not increasingly tax a single OS. This means parallel processing can scale to much higher levels than SMP alone.
Server Clustering Years ago, only a single processor was needed to run a server and
operate all its applications. With the advent of multiprocessing, two or more processors
shared a pool of memory, and the server could handle more and larger applications.
Later on, multiple servers were organized into groups with each server performing a
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 13
Chapter 1:
Network Hardware Concepts
13
specific task (i.e., file server, application server, and so on). Today, many high-end networks employ server clusters, where two or more server PCs act like a single server—
providing higher availability and performance than a single could handle. Applications
can move from one server to another, or run on several servers at once, and all transactions are transparent to the users.
Clustering provides higher availability and scalability than would be possible if the
computers worked separately. Each node in the cluster typically has its own resources
(processors, I/O, memory, OS, storage, and so on), and is responsible for its own set of
users. The high availability of a server cluster is provided by failover capability. When one
node fails, its resources can “failover” to one or more other nodes in the cluster. Once the
original node is restored to normal operation, its resources can be manually (or automatically) switched back. Server clusters are also easily scalable without an interruption
of service. Upgrades can be performed by proactively failing over the functions of a server
to others in the cluster, bringing that server down to add components, and then bringing
the server back up into the cluster and switching back its functions from the other servers.
Server clustering is not really a new idea, but they have generally been proprietary in
both hardware and software. IS managers are looking at clusters more seriously now as
they become more accessible using mass-produced, standards-based hardware like
RAID, SMP systems, network and I/O adapters, and other peripherals. While clusters are
poised to gain more sophistication in the future, a growing number of cluster options
are available today, and formal standards for clustering are still being developed.
Network Topology
In order to create a network, two or more PCs (and other peripheral devices) must be
connected together. However, there are several different ways to arrange these connections, and each connection scheme is known as a network topology. Each topology offers
its own unique capabilities and limitations. Unfortunately, topologies aren’t as simple
as plugging one computer into another; each topology requires certain cabling, NIC
adapters, network operating systems, and other devices. For example, a particular topology can determine not only the type of cable that is used, but also how the cabling runs
through floors, ceilings, and walls. While most network topologies use physical cables
to connect one computer to another, a growing number of networks use wireless transceivers for at least some connections. Topology can also determine how computers
communicate on the network. This part of the chapter introduces you to three traditional network topologies: bus, star, and ring.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 14
Server+ All-in-One Certification Exam Guide
14
Bus Topology
The bus is the simplest and most straightforward type of network topology, and is commonly used with Ethernet networks. With a bus (see Figure 1-1), computers are connected to each other in a straight line along a single main cable called a trunk (a.k.a.
backbone or segment). Bus networks are easy to connect and inexpensive to implement, and a computer failure won’t impair the entire network. However, overall bus
performance is limited, and cable breaks can shut down the entire network.
Bus Operation
Computers on a bus network communicate by addressing data to a particular computer
and sending out that data to all computers on the cable. Only the computer whose
address matches the address encoded in the original signal will accept the information;
all other computers simply ignore the data. Since data goes out to all computers simultaneously, only one computer at a time can send messages. As you might expect, this
also means the number of computers attached to the bus will affect network performance. The more computers there are on a bus, the more computers will be waiting to
put data on the bus, and the slower the network will be. Bus performance is difficult to
Terminator
Cable Segment
NIC
PC 1
NIC
PC 2
Terminator
NIC
PC 4
Figure 1-1 Typical bus topology
NIC
PC 3
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 15
Chapter 1:
Network Hardware Concepts
15
judge because numerous other factors affect bus performance (in addition to the number of computers) such as
• Hardware capabilities (i.e., NIC type) of computers on the network
• Total number of PCs waiting to be sent data (a.k.a. the network traffic)
• Types of applications (i.e., file system sharing) being run on the network
• Types of cable used on the network
• Distances between computers on the network (the overall length of the trunk)
The electronic signals that represent data are sent to the entire network, and travel
from one end of the cable to the other. If the signal is allowed to continue uninterrupted,
it will keep bouncing back and forth along the cable. This signal bounce can prevent other
computers from sending data. The signal must be stopped after it has reached the proper
destination address. To stop a signal from bouncing, a simple device called a terminator
is placed at each end of the network cable to absorb the signals. This clears the cable so
that other computers can send data. When using a bus topology, both ends of each cable
segment must be plugged into something. For example, a cable end can be plugged into
a computer or connector to extend the cable’s length. Any open cable ends not plugged
into something must be terminated to prevent signal bounce.
NOTE: Terminator problems are a common issue with bus networks, and they
should be checked and verified whenever network transmission problems arise.
Bus Disruptions
Computers on a bus topology will either transmit data to other computers or listen for
data from other computers on the network—they are not responsible for moving data
from one computer to the next. As a result, if one computer fails, it does not affect the
rest of the network. This is a main advantage of the bus topology. Unfortunately, bustype networks are extremely sensitive to cable breaks. A break in the cable will occur if
the cable is physically separated into two pieces (i.e., accidentally cut), or if at least one
end of the cable becomes disconnected (i.e., someone fiddles with a cable connection
behind the PC). In either case, one or both ends of the cable will not have a terminator,
the signal will bounce, and all network activity will stop, causing the network to go
down. The individual computers on the network will still be able to function as standalone PCs, but they will not be able to communicate with each other or access shared
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 16
Server+ All-in-One Certification Exam Guide
16
resources as long as the cable remains broken. The computers on the down segment
will continually attempt to establish a connection, and this will slow the workstations’
performance until the problem is resolved.
Expanding the Bus
It is fairly easy to expand the bus topology to accommodate more users and peripheral
devices. Simply remove a terminator from one end of the network trunk, add a cable to
another PC’s T connector, then replace the terminator at that last T connector (refer to
Figure 1-1). If you need to extend a given cable length to make it longer, you can fasten
two cable lengths together using a barrel connector. However, connections tend to
degrade signal strength, so it should be used only when absolutely necessary. Too many
connectors can prevent the signal from being received correctly. It’s preferable to
remove the shorter cable length and attach a more suitable length instead—one continuous cable is preferable to connecting several smaller ones with connectors. As an
alternative, a repeater can be used to connect two cable lengths. A repeater actually
boosts the signal strength, so the signal remains stronger across multiple connectors or
a longer piece of cable.
Star Topology
The star topology is slightly more sophisticated than a bus approach because all PCs on
the network are tied to a central connection point called a hub (see Figure 1-2). A star network is a bit more robust than the bus approach because connections are direct from the
PC to the hub. It’s an easy matter to add clients to the network simply by connecting
them to an available port in the hub (multiple hubs can be ganged together for larger networks with additional users). Because each connection is independent, you don’t need to
worry about terminators. A cable problem or PC fault will only affect that particular
workstation; it won’t disable the entire network. On the negative side, more cabling is
often required because each PC needs its own cable to the hub. Also, a hub failure can
disable all the PCs attached to it (though this is a fairly easy issue to troubleshoot).
Star Operation
Computers on a star network communicate by addressing data to a particular computer
and sending out that data through the hub to all computers on the network. Only the
computer whose address matches the address encoded in the original signal will accept
the information; all other computers simply ignore the data. Because data goes out to
all computers simultaneously, only one computer at a time can send messages. This
means the number of computers attached to the star will affect network performance.
The more computers there are on a star network, the more computers will be waiting to
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 17
Chapter 1:
Network Hardware Concepts
17
Cable segment
NIC
PC 1
NIC
Ports
PC 2
NIC
PC 3
Hub
NIC
PC 4
Figure 1-2 Typical star topology
send data to the hub and the slower the network will be. Star performance is difficult to
judge because numerous other factors affect network performance (in addition to the
number of computers), such as
• Hardware capabilities (i.e., NIC type) of computers on the network
• Performance and capabilities of the hub
• Total number of PCs waiting to be sent data (a.k.a. the network traffic)
• Types of applications (i.e., file system sharing) being run on the network
• Types of cable used on the network
• Distances between computers on the network
Unlike the bus topology, star network connections are not bothered by signal
bounce, so no special termination is needed. You simply connect the PC’s NIC adapter
port to the corresponding hub port.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 18
Server+ All-in-One Certification Exam Guide
18
Star Disruptions
Computers on a star topology will either transmit data through the hub to other computers or listen for data from the hub; they are not responsible for moving data from
one computer to another. As a result, if one computer fails, it does not affect the rest of
the network. This is an important advantage of the star topology. Also, the fact that all
of the network’s PCs must come together to a single point (the hub) means that the
hub(s), server(s), and other key network devices can all be conveniently located and
serviced in one place. This improves network troubleshooting and administration. If a
break or disconnection occurs with a cable, only that PC is effected, and the remainder
of the network can continue on normally. However, since the hub serves as a central
communication point in the star topology, a hub failure will quickly disable all the PCs
attached to it.
Expanding the Star
It is fairly easy to expand the star topology to accommodate more users and peripheral
devices. Additional users can simply be connected to an available port on an existing
hub. However, the added wiring becomes problematic. When a nearby PC is added to
a bus-type network, you only need to attach the new PC in-line with the existing trunk
wiring. When a new PC is added to a star-type network, you may need to run an entirely
new cable from the PC to the hub. This might require dozens (maybe hundreds) of feet
of additional wiring, which may need to be routed through floors, walls, and ceilings
depending on what’s between the user and the hub.
NOTE: In actual practice, network designers often mix topologies (i.e.,
bus/star) to make a more efficient use of cabling and equipment.
Ring Topology
The ring topology (usually called token ring) is a bit more sophisticated than a bus
approach because the trunk cable that connects all PCs on the network basically forms a
loop (see Figure 1-3). Computers are connected in a continuous network loop in which
a key piece of data (called a token) is passed from one computer to the next. The token is
a data frame (or packet) that is continuously passed around the ring. In actual practice,
token ring networks are physically implemented in a star configuration but managed logically as a loop. Workstations on a token ring network are attached to a specialized hub
called a multistation access unit (MAU). It’s an easy matter to add clients to the network
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 19
Chapter 1:
Network Hardware Concepts
19
Cable segment
Token
NIC
PC 1
NIC
Ports
PC 2
NIC
PC 3
MAU
NIC
PC 4
Figure 1-3 Typical ring (token ring) topology
simply by connecting them to an available port in the MAU (several MAUs can be ganged
together for larger networks with additional users). Since the overall effect is that of a
loop, you don’t need to worry about terminators. The token-passing approach ensures
that all PCs have equal access to the network, even when there are many users. On the
negative side, more cabling is often required because each PC needs its own cabling to the
MAU. Also, each computer must pass a token to the next, so a PC failure (or a MAU fault)
can impair the entire network. This can easily complicate the troubleshooting process.
Ring Operation
The most popular method of transmitting data around a ring is called token passing. The
token itself is little more than a short sequence of data bits that travel around a token
ring network, and each network has only one token. The token is passed (received and
retransmitted) from computer to computer. An advantage of this retransmission is that
each PC in the loop acts as a repeater—boosting the data signal to the next workstation.
This process of token passing continues until the token reaches a computer that has
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 20
Server+ All-in-One Certification Exam Guide
20
data to send. The sending computer modifies the token, puts an electronic address on
the data, and reinserts this new data package into the ring.
This data package passes by each computer until it finds the one with an address that
matches the address on the data. The receiving computer takes the data and attaches a
verification message to the token, which is re-addressed to the sender and returned to
the ring. The sending computer eventually receives the verification message, indicating
that the data has been received. After verification, the sending computer creates a new
token and inserts it on the network. The token continues to circulate within the ring
until another workstation needs it to send data. Token ring performance is difficult to
judge because numerous other factors affect network performance (in addition to the
number of computers), such as
• Hardware capabilities (i.e., NIC type) of computers in the ring
• Performance and capabilities of the MAU
• Total number of PCs waiting to be send data (a.k.a. the network traffic)
• Types of applications (i.e., file system sharing) being run on the network
• Types of cable used on the network
• Distances between computers on the network (the overall size of the ring)
Unlike the bus topology, ring network connections are not bothered by signal
bounce, so no special termination is needed. You simply connect the PC’s NIC adapter
port to the corresponding MAU port to add that PC to the loop.
Ring Disruptions
Computers in a token ring topology are constantly receiving and retransmitting tokens
from one computer to the next. As a result, if one computer fails or a cable breaks, it
interrupts the rest of the network. Since token rings also use MAUs to pass data from
one PC to the next (refer to Figure 1-3), a MAU failure can also disable the network.
These are important disadvantages of token ring topology, and can present a technician
with serious troubleshooting problems when faced with locating the break in a token
ring. On the plus side, a MAU provides a centralized communication point for network
administration and maintenance.
NOTE: The idea of a “ring” is only from a logical perspective. From a practical
standpoint, the network is wired as a “star”—an MAU is used to provide the
ring feature.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 21
Chapter 1:
Network Hardware Concepts
21
Expanding the Ring
Because a token ring is physically structured very similarly to a star network, it’s fairly
easy to expand the ring topology in order to accommodate more users and peripheral
devices. Additional users can simply be connected to an available port on an existing
MAU. As with star clients, however, the added wiring can be problematic. When a
nearby PC is added to a bus-type network, you only need to attach the new PC in-line
with the existing trunk wiring. When a new PC is added to a token ring network, you
might need to run an entirely new cable from the PC to the MAU. This may require
dozens (maybe hundreds) of feet of additional wiring, which may need to be routed
through floors, walls, and ceilings depending on what’s between the user and network’s MAU.
Network Hardware
Now that you’ve had a chance to learn about server types and network topologies, it’s
time to learn a bit more about the various hardware elements involved with the implementation of a network. Network hardware has a profound impact on the speed, quality, and overall performance of the network. For the purposes of this book, network
hardware includes hubs, repeaters, bridges, routers, gateways, network interface cards,
and cabling.
Hubs
Simply stated, a hub is a central connection device that joins computers in a star topography. A variation of the hub is a Multistation Access Unit (or MAU, sometimes called
a token ring hub) used to connect PCs in a token ring topology. Hubs are now standard
equipment in modern networks, and are typically classified as passive or active. A passive hub does not process data at all; it’s basically just a connection panel. By comparison, active hubs (sometimes called repeaters) regenerate the data in order to maintain
adequate signal strength. Some hubs also have the capability to handle additional tasks
such as bridging, routing, and switching. Hub-based systems are versatile and offer several advantages over systems that do not use hubs. For example, with an ordinary bus
topology, a break in the cable will take the network down. But with hubs, a break in any
of the cables attached to the hub affects only that limited segment of the network.
Most hubs are active—that is, they regenerate and retransmit signals in the same way
that a repeater does. Since hubs usually have eight to twelve ports for network computers to connect to, they are sometimes called multiport repeaters. Active hubs always
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 22
Server+ All-in-One Certification Exam Guide
22
require electrical power to run. Some hubs are passive (examples include wiring panels
or punch-down blocks). They act only as connection points, and do not amplify or
regenerate the signal. The signal just passes through the hub. Passive hubs do not
require electrical power to run. An emerging generation of hubs will accommodate several different types of cables. These are called hybrid hubs.
CAUTION: Be careful when connecting hubs. Crossover cables are wired differently than standard patch cables, and one will not work correctly in place of
the other. Check with the hub manufacturer to determine whether you need a
standard patch cable or a crossover cable.
Repeaters
As electrical signals travel along a cable, they degrade and become distorted. This effect
is called attenuation. As cable lengths increase, the effects of attenuation worsen. If a
cable is long enough, attenuation finally will make a signal unrecognizable, and this
will cause data errors in the network. Installing a repeater enables signals to travel farther by regenerating the network’s signals and sending them out again on other cable
lengths. The repeater takes a weak signal from one cable, regenerates it, and passes it to
the next cable. As you saw earlier, active hubs frequently act as repeaters, but standalone repeaters might be needed to support very long cable lengths.
NOTE: Broadband systems will use amplifiers rather than repeaters.
It is important to realize that repeaters are simply signal amplifiers. They do not
translate or filter the network signals from one cable to another. For a repeater to work
properly, both cables joined by the repeater must use the same packets, logical protocols, and access method. The two most common access methods are carrier sense multiple access with collision detection (CSMA/CD) and token passing. A repeater cannot
connect a segment using CSMA/CD to a segment using the token-passing access
method. In effect, a repeater will not allow an Ethernet network to talk to a token ring
network—there are other more sophisticated devices used for that type of translation.
However, repeaters can move packets from one kind of physical media to another. For
example, a repeater can take an Ethernet packet coming from a thin coaxial cable and
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 23
Chapter 1:
Network Hardware Concepts
23
pass it on to a fiber-optic cable (provided that the repeater is capable of accepting the
physical connections).
Because repeaters simply pass data back and forth between cables, you should realize that problem data (such as malformed packets) also will be processed by the
repeater. Bad data will not be filtered out, and excessive network traffic will not be managed. As a rule, avoid the use of repeaters when there is heavy network traffic or when
data filtering features are needed.
Bridges
A bridge offers more features for a busy network. A bridge can act like a repeater to
extend the effective length of a network cable. However, a bridge has more “intelligence,” and can also divide a network to isolate excessive traffic or problem data. For
example, if the volume of traffic from one or two computers (or a single department) is
flooding the network with data and slowing down the entire operation, a bridge could
isolate those computers (or department). Rather than distinguish between one protocol and another, bridges simply pass all protocols along the network. Because all protocols pass across bridges, it is up to the individual computers to determine which
protocols they can recognize. Bridges can also link different physical media such as
twisted-pair cable and thin coaxial cable.
Routing Data
A bridge also offers superior data-handling capabilities not provided by hubs and
repeaters. Bridges “listen” to all traffic, check the source and destination address of each
packet, and build a routing table (as information becomes available) so that they can
sort data to different parts of the network efficiently. Bridges actually have the capability to learn how to forward data. As traffic passes through the bridge, information about
the computer addresses is stored in the bridge’s memory. The bridge uses this information to build a routing table based on source addresses. Initially, the bridge’s memory
is empty and so is the routing table. As packets are transmitted, the source address is
copied to the routing table. With this address information, the bridge eventually learns
which computers are on which segment of the network.
When the bridge receives a packet, the source address is compared to the routing
table. If the source address is not there, it is added to the table. The bridge then compares the destination address with the routing table database. If the destination address
is in the routing table and is on the same network segment as the source address, the
packet is discarded (because it’s assumed that another PC on the same part of the network has received the data). This filtering helps reduce network traffic and isolate different parts of the network. If the destination address is in the routing table and not in
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 24
Server+ All-in-One Certification Exam Guide
24
the same segment as the source address, the bridge forwards the packet out of the
appropriate port to reach the destination address. If the destination address is not in
the routing table, the bridge forwards the packet to all its ports except the one on which
it originated.
Reducing Traffic
Remember that many PCs on a network may need to send data, but not all PCs may
need to receive that data. Often, all PCs must receive data to see whether the information is intended for that workstation, then each must wait for an opportunity to send
data itself. In a large network, this can significantly reduce network performance. However, large networks often group PCs into departments, and the data sent between
departments is often far less than the traffic sent between PCs within the same department. By using bridges to separate the overall company network into several smaller
departmental groups, it is possible to reduce the traffic going out to the entire network,
and thus improve the network’s overall performance.
Organizing Traffic with a Bridge
Let’s look at an example. Consider a company with five major departments: Sales,
Accounting, Shipping, Manufacturing, and Design. In an “open” network, traffic sent
from a PC in sales would eventually reach every other PC on the network (i.e., Accounting, Shipping, and so on). Since traffic from one department is most commonly
intended for other PCs in the same department, it’s often a waste of network time to
have all those other PCs check all that traffic. If a bridge is used to segregate the network
into five different areas, traffic sent from one PC in Design to another PC in Design
would not go out to the other areas of the network. This would reduce traffic because
all the other PCs would not need to check that traffic to see if it was intended for them.
If a PC in Design wanted to send traffic to another PC in Sales, the bridge would know
(through its routing table) which segment to relay that traffic to, and the other segments would not need to deal with that traffic. This controlling (or restricting) of the
flow of network traffic is known as segmenting network traffic. A large network is not limited to one bridge. Multiple bridges can be used to combine several small networks into
one large network.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 25
Chapter 1:
Network Hardware Concepts
25
Remote Connections
Bridges are often used to join smaller networks that are separated by large physical distances. For example, when two separate LANs are located at a great distance from each
other, they can be joined into a single network using two remote bridges connected
with synchronous modems to a dedicated data-grade telephone line.
Routers and Brouters
When you’re working in more complex network environments that use several different
network segments—each with different protocols and architectures—a bridge is often
inadequate to handle fast and efficient communication between diverse segments.
Such a complex network demands a sophisticated device that knows the address of
each segment, determines the best path for sending data, and filters broadcast traffic to
the local segment. This type of device is called a router. As with a bridge, routers can filter and isolate network traffic and also connect network segments. Further, routers can
switch and route packets across multiple networks. They do this by exchanging specific
protocol information between separate networks. Routers have access to more packet
information than bridges; routers use this additional information to improve packet
deliveries. Routers are used in complex networks because they provide better traffic
management. For example, routers can share status and routing information with one
another and use this information to bypass slow or malfunctioning connections.
There are two principal router types: static and dynamic. A static router is sometimes
called a manual router because all routes must be configured manually by the network
administrator. Routing tables are fixed, so the static router always uses the same route
(even if network activity changes). This means there’s no guarantee that the router is
using the shortest routes. By comparison, dynamic routers must be configured initially,
but they will adapt to changing network conditions automatically—using lower cost or
lower traffic routes as needed.
Routing Data
Routers maintain their own routing tables, which usually consist of network addresses
(though host addresses can also be kept if the network needs it). To determine the destination address for incoming data, the routing table includes all known network
addresses, logical instructions for connection to other networks, knowledge of the possible paths between routers, and even the costs of sending data over each path. Thus, a
router uses its routing table to select the best route for data transmission based on costs
and available paths. You should understand that the “routing tables” used for bridges
and routers are not the same thing. Routers require specific addresses. They understand
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 26
Server+ All-in-One Certification Exam Guide
26
only the network numbers that allow them to communicate with other routers and
local NIC addresses, so routers don’t talk to remote computers.
When routers receive packets destined for a remote network, they send them to the
router that manages the destination network. The use of routers allows designers to separate large networks into smaller ones and offers an element of security between the
segments. Unfortunately, routers must perform complex functions on each packet, so
they are slower than most bridges. For example, as packets are passed from router to
router, source and destination addresses are stripped off and then re-created. This
enables a router to route a packet from a TCP/IP Ethernet network to a server on a
TCP/IP Token Ring network—a feature unattainable with a bridge.
Reducing Traffic
Routers do not look at the destination node address. Instead, they look only at the network
address and will pass information only if the network address is known. Routers will not
allow corrupted (i.e., nonaddressed) data to be passed onto the network. This capability to
control the data passing through the router reduces the amount of traffic between networks
and allows routers to use these links more efficiently than bridges. Consequently, routers can
greatly reduce the amount of traffic on the network and the wait time experienced by users.
Remember that not all protocols are routable (you’ll see more about protocols in Chapter
2). Typical routable protocols include DECnet, Internet Protocol (IP), and Internetwork
Packet Exchange (IPX), while protocols such as Local Area Transport Protocol (LATP) or
NetBIOS Extended User Interface (NetBEUI) are not routable. Routers are available that can
accommodate multiple protocols (such as IP and DECnet) in the same network.
Selecting a Route
One distinct advantage enjoyed by routers is that they can support numerous active
paths between LAN segments and select redundant paths if necessary. Since routers can
link segments that use completely different data packaging and access schemes, there
are usually several possible paths available for a router to use. For example, if one router
does not function, data can still be sent using alternative routes. This also applies to
network traffic. If one path is very busy, the router identifies an alternative path and
sends data over that one instead. Routers use powerful algorithms such as OSPF (Open
Shortest Path First), RIP (Routing Internet Protocol), or NLSP (NetWare Link Services
Protocol) to determine an appropriate transmission path for a data packet.
Brouters
The functional distinctions between bridges and routers are blurring as technology
advances. Some bridges have advanced intelligence that allows them to handle tasks
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 27
Chapter 1:
Network Hardware Concepts
27
that would normally require a router. These advanced bridges are called brouters. A
brouter can act as a “router” for one protocol and “bridge” for all the others. A brouter
can route selected routable protocols, bridge nonroutable protocols, and provide more
cost-effective and manageable internetworking than separate bridges and routers.
Gateways
A gateway acts as a powerful interpreter designed to connect radically different networks.
Although slower than a bridge or router, a gateway can perform complex functions such
as translating between networks that speak different languages (using techniques such
as protocol and bandwidth conversion). For example, a gateway can convert a TCP/IP
packet to a NetWare IPX packet (and vice versa). Gateways enable communication
between entirely different architectures and environments. They effectively repackage
and convert data going from one type of network to another so that each can understand the other’s data. A gateway repackages information to match the requirements of
the destination system, and changes the format of a message so that it conforms to the
application running at the receiving end of the transfer. In most cases, gateways are taskspecific, which means that they are dedicated to a particular type of transfer. They are
often referred to by their task (i.e., Windows NT Server-to-SNA gateway).
Network Interface Cards (NICs)
NICs (also known as LAN adapters) function as an interface between the individual
computer (server or client) and the network cabling (see Figure 1-4). Internally, the NIC
must identify the PC on the network and buffer data between the computer and the
cable. When sending data, the NIC must convert the data from parallel bytes into serial
bits (then back again during reception). On the network side, an NIC must generate the
electrical signals that travel over the network, manage access to the network, and make
the physical connection to the cable. Every computer on the network must have at least
one NIC port installed. Modern NICs increase their effective throughput using
advanced techniques of adapter teaming such as adapter fault tolerance (AFT), which provides automatic redundancy for your adapter. If the primary adapter fails, the secondary
takes over. Adaptive load balancing (ALB) allows balancing the transmission data flow
between two to four adapters. You’ll see much more about NICs in Chapter 10.
Cabling
Finally, networks of all sizes and configurations depend on the physical cabling that
connects all the PCs and other hardware together. Cabling (also referred to as network
media) comes in many different configurations, but common cabling used for everyday
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 28
Server+ All-in-One Certification Exam Guide
28
Figure 1-4 The Symbios SYM22915 network interface card (Courtesy of LSI Logic Corp.)
networking includes unshielded twisted pair (UTP), coaxial cable, shielded twisted pair
(STP), and fiber-optic (FO) cable. As a technician, you should understand the three
main considerations for cabling:
• Resistance to crosstalk (electrical currents between pairs of wires in the same cable)
• Resistance to interference from outside electrical fields (noise created by electric
motors, power lines, relays, and transmitters)
• Ease of installation
These are important issues because cables resistant to crosstalk and interference can
be run longer and support higher data transmission rates. For example, coaxial and
shielded twisted-pair cable have a thin metal foil outer layer that offers good resistance
to electrical noise, but the extra foil creates a larger, thicker cable that is more difficult
to pull through conduit and walls during installation. Unshielded twisted pair is thinner and easier to install, but offers less resistance to electrical noise. By comparison,
fiber-optic cable carries light signals instead of electrical pulses, so it is impervious to
electrical interference. This allows fiber-optic cable to carry signals faster and farther
than any other type of cable. Unfortunately, FO cable is often far more expensive than
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 29
Chapter 1:
Network Hardware Concepts
29
other cable types, and proper installation demands specialized tools and training. The
following section explains these major cable types in detail.
Understanding Network Media
Every computer in any kind of network must ultimately be connected to one another.
These connections are responsible for transmitting vast amounts of information
between the computers and peripheral devices. Although wireless networking is growing in popularity, the vast majority of network connections are made physically using a
variety of cable types—each intended for a specific type of network architecture. We
usually refer to this interconnecting wiring as network media. While there are more than
2,000 different types of cabling, most network applications use only three different
cable types: coaxial, twisted pair, and fiber-optic cable.
Signal Transmission
In order to understand the importance and characteristics of cabling, you should
understand the idea of “bandwidth,” and the two approaches used to transmit data
signals: baseband and broadband. Bandwidth is simply the amount of data that can be
handled by a cable or device over a given time (sometimes called a data transfer rate or
transmission rate). For example, coaxial cable can typically support from 4Mbps–
100Mbps (depending on the quality and length of the cable), while fiber-optic cable
can handle up to 1Gbps. Of course, the network architecture and network interface
cards must also support the same bandwidth if you’re going to use the cabling to its
optimum capacity.
Baseband transmission employs digital signaling to use the entire bandwidth of the
cable. The serial cable that connects your PC’s COM port to an external modem is a very
simple example of baseband transmission. Twisted pair and fiber-optic cables are typically used in baseband systems, though coaxial cables are sometimes used that way
also. Baseband systems normally use repeaters to receive incoming signals and retransmit them at their original strength to increase the practical length of a cable.
By comparison, broadband transmission uses analog signaling across a wide range of
frequencies. Your everyday cable service that brings 200 channels to your TV and highspeed Internet access to your PC is an example of broadband transmission. With broadband, each signal is allocated a part of the total bandwidth. Every device associated
with the system (i.e., all computers connected to the LAN cable) must then be tuned so
that they use only the frequencies that are within the allocated range. Broadband systems use amplifiers (rather than repeaters) to regenerate analog signals to their original
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 30
Server+ All-in-One Certification Exam Guide
30
strength. As with baseband signals, broadband signals flow in one direction only, so
there must be two paths for bi-directional data. The obvious solution is to use two
cables: one for transmission and the other for reception. However, since broadband
also allows numerous signals on the same cable, bandwidth is often split, with half of
the bandwidth assigned to “transmit” channels and half of the bandwidth allocated to
“receive” channels.
Coaxial Cable
Coaxial cable (coax) is an inexpensive, flexible, and rugged type of transmission cable.
Coaxial cables use a single copper wire at the center of an internal insulating layer, covered by a finely braided metal shield, and covered by a protective outer jacket (see Figure 1-5). Its light weight and flexibility make coaxial cable easy to install in a wide range
of office environments. That wire in the middle of the coaxial cable is what actually carries the signal. It is often a solid copper wire, but might sometimes be stranded
aluminum. A fairly thick dielectric insulating layer surrounds the core, and this separates the core from the metal shielding. A braided wire mesh acts as an electrical ground
and protects the core from electrical noise and crosstalk. The shielding also protects
transmitted data from electrical noise. For additional protection, a coaxial cable may
incorporate one layer of foil insulation and one layer of braided metal shielding (dual
shielding), or two layers of foil insulation and two layers of braided metal shielding
(quad shielded). Additional shielding adds greatly to the cable’s cost and weight.
Finally, a protective outer cover of rubber, Teflon, or PVC plastic is used to jacket the
cable. You’ll generally find two types of coaxial cable used in networking: thin and thick.
All coaxial cables are attached using specialized quick-twist connectors called BNC
connectors. A BNC T connector is an adapter used to attach two lengths of cable to your
Outer Jacket
Braided Shield
Insulator
Core
Figure 1-5 Diagram of a typical coaxial cable
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 31
Chapter 1:
Network Hardware Concepts
31
NIC. If you need to adapt two lengths of cable to make one longer run, use a BNC barrel connector. Finally, you’ll need a BNC terminator to cap each end of the cable run
(usually attached to the unused port of the last BNC T connectors).
Thinnet Cable
As the name implies, thinnet cable is thin—roughly 0.25 in. (diameter)—and can carry
electrical signals for over 600 ft. The cable industry refers to this common type of cable
as RG-58. Thinnet cable presents a 50 impedance (signal resistance) to the data signals flowing through it. The cable’s small diameter makes it flexible and easy to install
just about anywhere.
Thicknet Cable
Thicknet cable (sometimes called standard Ethernet cable because of its use with early
Ethernet networks) offers a diameter of 0.5 in.—twice the diameter of thinnet cable. The
copper core wire is also thicker, and this allows thicknet cable to transfer signals well over
1,500 ft. This capability to carry signals a great distance makes thicknet an ideal choice for
a backbone cable that’s capable of connecting several smaller thinnet network segments.
Unfortunately, thicknet cable does not bend easily, so it is considerably harder to install.
The transition from thicknet to thinnet cable is made with a transceiver device. The
transceiver’s sharp points pierce the thicknet cable (referred to as a vampire tap) in order
to contact the cable’s core and shielding. An output cable from the transceiver attaches
to the computer’s corresponding NIC port. In many cases, the NIC adapter requires an
attachment unit interface (AUI) port connector (also known as a Digital Intel Xerox
[DIX] connector) to accommodate the transceiver.
Cable Grades
Chances are that you’ll be running coaxial cable through walls, in ceilings, under floors,
and in or through other odd locations throughout your office. It’s important to remember that ordinary coaxial cable uses a jacket of PVC or other synthetic material that
makes it easy to pull and route. However, building fire codes generally prohibit the use
of everyday coaxial cable in a building’s plenum (the shallow space in many buildings
between the false ceiling and the floor above). During a fire, PVC jackets will burn and
generate poisonous gases. Coaxial cable rated for plenum-grade use employs insulation
and jacket materials that are certified to be fire resistant and produce a minimum
amount of smoke. This reduces poisonous chemical fumes in the event of a fire.
Plenum cable can also be used in the plenum area and in vertical runs (i.e., up a wall)
without conduit. Be sure to review and understand the fire safety codes for your location when building, servicing, or expanding your network.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 32
Server+ All-in-One Certification Exam Guide
32
Twisted-Pair Cable
Another popular cable type that is commonly used with current networks is called
twisted pair. As the name suggests, a twisted pair is little more than two insulated lengths
of copper wire twisted around each other—though a typical twisted pair cable carries
two, three, or even four pairs of wire contained in a single plastic, PVC, or Teflon jacket
(see Figure 1-6). The physical twisting of the wires works to cancel out electrical noise
from adjacent pairs, as well as other noise sources such as motors, relays, and transformers. Twisted-pair cable is either shielded or unshielded, and the choice between
these two may have a profound impact on the reliability of your data (especially if you
must carry data over a distance).
Twisted-pair cabling uses RJ-45 telephone connectors. At first glance, these connectors look like the RJ-11 telephone connectors that attach your telephone cord to the wall.
The RJ-45 connector is slightly larger, and will not fit into an RJ-11 telephone jack. The
RJ-45 connector handles eight cable connections, while the RJ-11 supports only four.
This means you can’t accidentally exchange your telephone and network connectors.
Unshielded Twisted Pair (UTP)
When there are one or more pairs of twisted wire, but none of the pairs (nor the full
cable) contain additional metal foil or braid for shielding, the twisted-pair cable is said
to be unshielded twisted pair (UTP). UTP is an inexpensive and versatile cable that has
RJ-45 Connector
Twisted Pair Cable
Four Pairs
Figure 1-6 A typical twisted pair cable
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 33
Chapter 1:
Network Hardware Concepts
33
been popular with 10BaseT networks. The maximum cable length for a UTP network
segment is about 328 ft. National standards organizations have specified the type of
UTP cable that is to be used in a variety of building and wiring situations. These standards include five distinct categories for UTP:
• Category 1 This is traditional UTP telephone cable that was intended to carry
voice but not data. Most telephone cable prior to 1983 was Category 1 cable.
• Category 2 This type of UTP cable is designed for data transmissions up to
4Mbps (megabits per second) with four twisted pairs of copper wire.
• Category 3 This category includes UTP cable for data transmissions up to
16Mbps with four twisted pairs of copper wire.
• Category 4 This is UTP cable intended for medium-speed data transmissions up
to 20Mbps with four twisted pairs of copper wire.
• Category 5 This category certifies UTP cable for high-speed data transmissions
up to 100Mbps with four twisted pairs of copper wire.
One reason why UTP is so popular is because many buildings are prewired for
twisted-pair telephone systems using a type of UTP. In fact, extra UTP is often installed
to meet future cabling needs as part of the facility’s prewiring process. If preinstalled
twisted-pair cable meets the category requirements to support data transmission, it can
be used in a computer network directly. However, common telephone wire (Category 1
wire) might not have the twisting and other electrical characteristics required for clean,
secure, computer data transmission.
Shielded Twisted Pair (STP)
In order to avoid degradation of the data because of crosstalk and noise, twisted pairs
of wire are often shielded with a wrap of thin metal foil. A fine copper braid then surrounds all the pairs, and a thick protective jacket of plastic, Teflon, or PVC is applied.
These metal shields reduce signal errors and allow the cable to carry data faster over a
greater distance. Other than the shielding, UTP and STP cable is identical.
Fiber-Optic Cable
Traditional wire cable carries data in the form of electrical signals (i.e., voltage and current). Fiber-optic cable is fundamentally different in that it uses specialized optical
materials to carry data as pulses of light. This makes fiber-optic cable uniquely immune
to electrical noise and crosstalk, and allows FO cable to carry a high data bandwidth
over several miles with surprisingly effective security—that is, the FO cable cannot be
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 34
Server+ All-in-One Certification Exam Guide
34
tapped without interrupting the data. Fiber-optic cable transmissions are extremely
fast, easily handling 100Mbps, and with demonstrated data rates to 1Gbps.
An optical fiber consists of an extremely thin fiber of glass (called the core) surrounded by another layer of glass with slightly different optical characteristics (known
as the cladding). The cladding effectively keeps light signals in the core material as it
passes down the cable. Since each fiber only passes signals in one direction, a complete
cable includes two strands in separate jackets: one strand transmits and the other
receives. A coating of plastic surrounds each glass strand, and Kevlar fibers provide
strength. Plastic (rather than glass) is sometimes used as the optical material because it
is cheaper and easier to install, but plastic is not as optically clear as glass and cannot
carry light signals over the same long distance.
IBM Cabling System
If you work with network cabling for any length of time, chances are that you’ll
encounter the IBM cabling system. IBM introduced its cabling system in 1984 to ensure
that network cabling and connectors would meet the specifications of their own equipment. The IBM cabling system classifies cable into “types” rather than categories. For
example, Category 3 cable (voice-grade UTP cable) is denoted as Type 3 cable in the
IBM system. The major types of IBM cabling are listed here:
• Type 1 This is standard shielded twisted-pair (STP) cable used for computers
and multistation access units (MAUs).
• Type 2
This is considered STP voice and data cable.
• Type 3
This is conventional UTP voice-grade cable.
• Type 4
Undefined.
• Type 5
This is industry-standard fiber-optic (FO) cable.
• Type 6
This is STP cable used for data patch applications.
• Type 7
Undefined.
• Type 8 This is referred to as carpet cable—STP cable housed in a flat jacket for
use under carpets.
• Type 9
This is plenum-grade (fire safe) STP cable.
One element unique to the IBM cabling system is the cable connector. These IBM
Type A connectors (commonly known as universal data connectors) are different from
standard BNC or other connectors. They are neither male nor female—you can connect
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 35
Chapter 1:
Network Hardware Concepts
35
one to another by flipping either one over. These IBM connectors require special faceplates and distribution panels to accommodate their unique shape.
Network Architectures
The architecture of a network is basically the way it is designed, and the way information
is exchanged. There are three basic types of network architectures that you should be
familiar with: Ethernet, token ring, and ARCnet. This part of the chapter examines these
architectures in more detail and explains the impact of cabling and access techniques.
Understanding the Packet
To the novice, it might seem that networks exchange information as a continuous
stream of data between computers. This is not the case. Sending large amounts of data
at one time causes other computers to wait idly while the data is being moved. This
monopolizes the network and wastes the time of other users waiting to use the network,
especially if a transmission error requires the data to be retransmitted. Rather than
exchange entire files at one time, data is broken down into much smaller chunks. Each
chunk is wrapped with the essential details needed to get the data to its correct destination without errors. These organized chunks are called packets (or frames), and may
require many packets to transfer an entire file from one network computer to another.
By transferring data in small packets, wait times seen by other computers on the network are much shorter because numerous computers on the network take turns sending packets. Should a packet arrive at a destination computer in a damaged or
unreadable state (because of signal attenuation), it is much easier and faster to retransmit that packet rather than the entire file. Packet data typically contains information
(such as e-mail messages or files), but many other types of data can be exchanged in
packets, such as command and control data or session control codes (i.e., feedback that
indicates a packet was received properly, or requires retransmission).
Packet Organization
A packet is basically made up of three parts: header, data, and trailer. Data is preceded
by a header, which includes a signal that indicates a packet is being transmitted, a
source address, a destination address, and clock information to synchronize the transmission. The actual data being sent is included after the header. The header of the
packet may vary greatly in size depending on the particular network, but most networks
include from 512 bytes to 4KB. Remember that most files are much larger than this, so
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 36
Server+ All-in-One Certification Exam Guide
36
it may take many packets to transmit a complete file. A trailer follows the data. The exact
content of a trailer may vary, but a trailer usually contains error-checking information
called a cyclical redundancy check (CRC). The CRC is a number produced by a mathematical calculation performed on the packet at its source. When the packet arrives at its
destination, the calculation is made again. If the results of both calculations are the
same, the data in the packet has remained intact. If the calculation at the destination
differs from that at the source, the data has changed during the transmission, and a
retransmission is requested.
NOTE: The exact formation and length of a packet will depend on the network’s communication protocol—the set of rules or standards that enable
computers to connect with one another and exchange information with as
little error as possible.
Understanding Access Methods
Of course, the computers on a network can’t just start spewing packets at any point.
While network traffic may seem to be moving simultaneously, a closer look will reveal
that computers are actually taking turns placing their data on the network. If two computers place their data on the network at the same time, both data packets would “collide” and be destroyed. The flow of network traffic must be carefully regulated. The
rules that govern how data is sent onto (or taken from) a network are called the access
method. An access method provides the traffic control needed to organize data transmissions on the network. It is also important to realize that all computers on the network must use the same access method. Otherwise, network problems would occur
because some access methods would monopolize the cable. There are three major
access methods: CSMA, token passing, and demand priority.
CSMA/CD
In the carrier sense multiple access with collision detection method (CSMA/CD), each
computer on the network (clients and servers alike) checks to see that the cable is free
before it sends a packet. If data is currently on the cable, the computer will not send; it
will wait and check the cable again. Once a computer has transmitted data on the cable,
no other computer can transmit data until the original data has reached its destination
and the cable is free again. This is often known as a contention method because two or
more computers are contending for the network cable. If two or more computers happen
to send data at exactly the same time, there will be a data collision. The two computers
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 37
Chapter 1:
Network Hardware Concepts
37
involved will stop transmitting for random periods of time, and then attempt to retransmit. The CSMA/CD technique is only useful up to about 1.5 miles. Beyond that, it might
not be possible for a computer at one end to sense that a computer at the other end is
transmitting. CSMA/CD can be frustratingly slow when network traffic is heavy.
CSMA/CA
The carrier sense multiple access with collision avoidance method (CSMA/CA) is similar to CSMA/CD, but allows each computer to signal its intention to transmit data
before the packet is actually sent. This enables other computers to sense when a data
collision might occur, and thus avoid transmissions that might result in collisions. The
problem with this approach is that broadcasting the intent to transmit actually adds to
the network traffic and can result in even slower network performance. This makes
CSMA/CA the least popular access method.
Token Passing
With the token-passing method, a special type of packet (called a token) is circulated
around a cable ring from computer to computer. In order for any computer on the ring
to send data across the network, it must wait for a free token. When a free token is
detected, the computer waiting for the token will take control of it. The sending computer then modifies the packet to include appropriate headers, data, and trailers, and
sends the new packet on its way. The receiving computer accepts the packet and its data,
and then creates another token for the sending computer indicating that the packet has
been received. When the sending computer receives this token, it creates a new free
token and passes it back onto the ring. When a token is in use by a computer, other
computers cannot transmit data. Because only one computer at a time can use the
token, no contention (or collision) takes place, and no time is spent waiting for computers to re-send tokens because of network traffic.
Demand Priority
The demand-priority method is a fairly new approach intended to service the 100Mbps
Ethernet standard (IEEE 802.12 or 100VG-AnyLAN) based on the star (or star/bus)
topology. Hubs manage network access by doing round-robin searches for requests to
send from all nodes on the network. As with CSMA/CD, two computers using demand
priority can cause contention by transmitting at exactly the same time. With demand
priority, however, it is possible to decide which types of data will be given priority if
contention occurs. If a hub receives two requests at the same time, the highest-priority
request is serviced first. If the two requests are of the same priority, both requests are
serviced by alternating between the two.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 38
Server+ All-in-One Certification Exam Guide
38
Demand priority offers several powerful advantages over CSMA/CD. First, communication only takes place between the sending computer, hub, and destination computer. This means transmissions are not broadcast to the entire network. Second,
demand priority uses twisted-pair cabling (four pairs), which allows computers on the
network to receive and transmit at the same.
Ethernet
Ethernet can trace its origins back to the late 1960s when the University of Hawaii
developed a network that would connect computers across its large campus. This early
network employed a bus topology, baseband transmission, and a CSMA/CD access
method. Xerox built upon this scheme, and by 1975 introduced the first Ethernet networking products intended to operate over 2.5Mbps and connect more than 100 computers across a 1km trunk. This early implementation of Ethernet proved so popular
that Xerox, Intel, and Digital (DEC) collaborated on the 10Mbps Ethernet standard
(now one of several specifications allowing computers and data systems to connect and
share cabling). Ethernet has become one of the most popular network architectures for
the desktop computer, and is used in network environments of all sizes. Today, Ethernet is considered to be a nonproprietary industry standard that is widely supported by
network hardware manufacturers. Ethernet can be summarized as follows:
• Topologies Bus or star
• Transmission Baseband
• Access method CSMA/CD
• Bandwidth 10Mbps or 100Mbps
• Cable type(s)
Coaxial (thicknet or thinnet) or UTP
• IEEE specification IEEE 802.3
Ethernet Packets
An Ethernet packet (commonly called a frame among Ethernet users) is between 64
and 1,518 bytes long (512–12,144 bits), and every packet includes control information.
For example, the Ethernet II packet format used for Transmission Control
Protocol/Internet Protocol (TCP/IP) is the standard for data transmission over networks (including the Internet). This packet includes six distinct areas. The preamble
marks the start of the packet (similar to the start bit used in serial communication). The
addresses denote the destination and source addresses for the packet. A type entry is used
to identify the network layer protocol—usually either IP (Internet Protocol) or IPX
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 39
Chapter 1:
Network Hardware Concepts
39
(Novell’s Internetwork Packet Exchange). The packet’s data then follows, and the packet
is concluded by error checking (CRC) information.
Ethernet Performance Notes
Ethernet performance can be improved by dividing a crowded segment into two lesspopulated segments then joining them with either a bridge or a router. This reduces the
traffic on each segment; because fewer computers are attempting to transmit onto the
segment, the apparent access time improves. You might consider dividing segments if
new users are quickly joining the network or if new bandwidth-intensive applications
(i.e., database or video software) are added to the network.
Ethernet architecture is also quite versatile, and can use multiple communication
protocols or connect mixed computing environments such as NetWare, UNIX, Windows, or Macintosh. Ethernet will work with most popular network operating systems,
including
• Microsoft Windows 95, Windows 98, and Windows 2000
• Microsoft Windows NT Workstation and Windows NT Server
• Microsoft Windows 2000 Professional and Windows 2000 Server
• Microsoft LAN Manager
• Microsoft Windows for Workgroups
• Novell NetWare
• IBM LAN Server
• AppleShare
• UNIX
10BaseT (IEEE 802.3)
10BaseT is an Ethernet standard designed to support 10-Mbps baseband data transmission over Category 3, 4, or 5 twisted-pair cable (UTP). UTP cable is more common, but
STP can be substituted without difficulty. Cables are connected with RJ-45 connectors.
Each computer uses two pairs of wire: one pair to receive and the other pair to transmit.
While Ethernet LANs are traditionally configured in a bus topology, a growing number
are set up as a star topology (using bus signaling and access methods). The hub of a
10BaseT network typically serves as a multiport repeater. The maximum length of a
10BaseT segment is 328 ft, though repeaters can be used to extend this maximum
length. The minimum cable length between computers is 8 ft. A 10BaseT Ethernet LAN
can serve up to 1,024 computers.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 40
Server+ All-in-One Certification Exam Guide
40
NOTE: Although twisted-pair cable is used to connect computers to a
hub, coaxial cable or fiber-optic cable may serve as a backbone between
10BaseT hubs.
10Base2 (IEEE 802.3)
10Base2 is an Ethernet standard designed to support 10Mbps baseband data transmission over thin coaxial (thinnet) cable. Cables are connected with BNC connectors
(including barrel connectors, T connectors, and terminators). 10Base2 Ethernet LANs
are traditionally configured in a bus topology. The maximum length of a 10Base2 segment is 607 ft., though repeaters can join up to five segments to create an effective bus
length of over 3,000 ft. The minimum cable length between computers is 2 ft. A
10Base2 Ethernet LAN will only serve up to 30 computers per segment, but this is often
ideal for small department and workgroup situations.
NOTE: BNC barrel connectors can be used to connect thinnet cable lengths
together. However, the use of barrel connectors should be kept to a minimum
because each connection in the cable reduces the signal quality and adds to the
danger of cable separation and accidental disconnection, which can effectively
shut down the network.
10Base5 (IEEE 802.3)
10Base5 is an Ethernet scheme (called standard Ethernet) designed to support 10Mbps
baseband data transmission over thick coaxial (thicknet) cable. 10Base5 Ethernet
LANs are traditionally configured in a bus topology, and the maximum length of a
10Base5 segment is 1,640 ft., though repeaters can join up to five segments to create
an effective bus length of over 8,200 ft. The backbone (or trunk) segment is the main
cable from which transceiver cables are connected to stations and repeaters. The minimum cable length between transceivers is 8 ft. A 10Base5 Ethernet LAN will only serve
up to 100 computers per segment, and this is often ideal for small to mid-sized network situations.
Cabling a 10Base 5 network can be a bit more involved than other Ethernet configurations. The thicknet cabling includes transceivers that provide communications
between the computer and the main LAN cable, and are attached to the main cable with
vampire taps. Once a transceiver is placed on the main cable, a transceiver cable (a.k.a.,
a drop cable) connects the transceiver to the NIC. A transceiver cable attaches to an NIC
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 41
Chapter 1:
Network Hardware Concepts
41
through an AUI (or DIX) connector. Other cabling is attached with N-series connectors,
including barrel connectors and terminators.
NOTE: A thicknet network can combine as many as five cable segments
connected by four repeaters, but only three segments can have computers
attached. This means two segments are untapped and often known as
“inter-repeater links.” This is known as the 5-4-3 rule. Remember that the
length of the transceiver cables is not used to measure the distance of
the thicknet cable—only the end-to-end length of the thicknet cable segment
itself is used.
10BaseFL
It is also possible to run an Ethernet network over fiber-optic cable. 10BaseFL is
designed to support 10Mbps baseband data transmission over fiber-optic cable
between computers and repeaters. The main reason for using 10BaseFL is to accommodate long cable runs between repeaters, such as between buildings. The maximum distance for a 10BaseFL segment is about 6,500 ft.
100BaseVG
Originally developed by Hewlett-Packard, the 100BaseVG (a.k.a. voice grade) AnyLAN
scheme is an emerging networking technology that combines elements of both Ethernet and token ring architectures. This type of architecture is known by several terms:
100VG, AnyLAN, 100BaseVG, or simply VG. 100BaseVG supports a minimum data rate
of 100Mbps in a star (or cascaded star) topology across Category 3, 4, and 5 twistedpair (as well as fiber-optic) cable. Because 100BaseVG is compatible with existing
10BaseT cabling systems, it is a simple matter to upgrade from existing 10BaseT installations (though new hubs and NIC adapters will be required). 100BaseVG uses the
demand-priority access method that allows for two priority levels (low and high), and
supports both Ethernet frames and token ring packets. While data transmission rates
are higher, the cable distances of 100BaseVG are limited when compared to other
implementations of Ethernet. A cable run from the 100BaseVG hub to a computer cannot exceed about 820 ft.
100Base“X”
There are several variations of the 100Base”X” family depending on the media being
used. 100BaseT4 uses four-pair Category 3, 4, or 5 UTP cable; 100BaseTX uses two-pair
Category 5 UTP or STP cable; and 100BaseFX uses two-strand fiber-optic cable. But all
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 42
Server+ All-in-One Certification Exam Guide
42
are referred to as Fast Ethernet because of their 100-Mbps transmission speeds.
100Base”X” also uses CSMA/CD in a star-wired bus topology (similar to 10BaseT where
all cables are attached to a hub).
Token Ring
IBM introduced the token ring architecture in 1984 for personal, midrange, and mainframe (i.e., SNA) computers. The main objective behind token ring was to establish a
simple and reliable wiring method using twisted-pair cable, which could connect individual workstations to a central location. The architecture of a token ring network is
technically a physical ring. However, rather than cabling the network PCs in an actual
circle (which could make upgrades and workstation additions a real nightmare), the
token ring approach uses a star topology where all PCs are connected to a central hub
called a multistation access unit (MAU). In effect, the ring is provided by the MAU rather
than by the physical cabling.
Cable segments can range from 148 ft–656 ft (depending whether the cable is
shielded or unshielded), and requires a minimum of 8 ft between computers. A segment will support up to 72 computers using unshielded cable, though up to 260 computers can be supported on a segment with shielded cable. Rings can be connected
through the use of bridges. Although Ethernet is more popular, many large companies
are selecting token ring architecture to support mission-critical applications. Token ring
architectures can be summarized as follows:
• Topologies Star
• Transmission Baseband
• Access method Token passing
• Bandwidth 4Mbps or 16Mbps
• Cable type(s)
Shielded and unshielded twisted pair (IBM Types 1, 2, and 3)
• IEEE specification IEEE 802.5
Token Ring Packets
The token ring packet is a bit more involved than an Ethernet packet, but contains the
same essential information. A start delimiter indicates the start of a packet, and access
control information describes the packet as a token (being passed around the network)
or data (having a specific destination). Packet control information will carry details for
all computers or only for one computer. The packet is directed with a destination address
and source address, and then the data to be transferred is included. Data might also
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 43
Chapter 1:
Network Hardware Concepts
43
include network commands or status information. A packet check sequence will provide
CRC error-checking information, and an end delimiter marks the end of the packet.
Packet status information is tagged onto the packet that tells whether the packet was recognized or copied, or if the destination address was even available. This information is
passed back to the sending computer.
Token Ring Operation
Now is a good time to review token ring operation. When the network initializes, a
token is generated that then travels around the ring and polls each computer until one
of the computers wants to transmit data—that computer then takes control of the
token. After a computer captures the token, it sends a data packet out to the network.
The packet proceeds around the ring until it reaches the computer with the address that
matches the destination in that packet. The destination computer copies the frame into
a receive buffer and updates the packet’s Packet Status field to indicate that the information was received. The updated packet continues around the ring until it arrives back
at the sending computer. The sending computer acknowledges the successful transmission, and then removes the packet from the ring and transmits a new token back to the
ring. It is important to remember that a computer cannot transmit unless it has possession of the token, and no other computer can transmit data while the token is in use by
a computer. Only one token at a time can be active on the network, and the token can
travel in only one direction around the ring.
System Monitoring and Fault Tolerance
One major advantage of the token ring architecture is its self-monitoring (or self-diagnosing) capability. Normally, the first computer to come online in a token ring network
is assigned to monitor network activity. The monitoring computer verifies that packets
are being delivered and received correctly by checking for packets that have circulated
the ring more than once (and ensuring that only one token is on the network at a time).
This monitoring process is called beaconing. A beacon announcement is produced every
seven seconds. The beacon is passed from computer to computer throughout the entire
ring. If a station does not receive an expected announcement from a PC upstream, it
tries to notify the network. It sends a message of the neighbor that did not respond and
attempts to diagnose the problem without disrupting the entire network. If a correction
cannot be made automatically, service will be required.
In addition, MAUs incorporate a certain amount of fault tolerance. When one computer fails in a “true” token passing network, the token cannot be passed and the network fails. MAUs are designed to detect a NIC failure and disconnect that computer
from the network. This bypasses the failed PC so that the token can continue on to the
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 44
Server+ All-in-One Certification Exam Guide
44
next subsequent computer. This means a faulty computer or connection will not affect
the rest of the token ring network.
ARCnet
ARCnet (Attached Resource Computer Network) is a simple, inexpensive, flexible
scheme designed for workgroup-size networks. Datapoint Corporation developed the
ARCnet network architecture in 1977, and the first ARCnet cards shipped in 1983. ARCnet uses a token-passing access method in a star-bus topology, passing data packets
with up to 508 bytes of data at rates approaching 2.5Mbps. ARCnet Plus (a successor to
the original ARCNet) uses data packets with up to 4,096 bytes and supports data transmission rates up to 20Mbps.
As a star topology, each computer is connected by cable to a hub. The standard
cabling used for ARCnet is 93 RG-62 coaxial cable, though RG-58 coaxial, twistedpair, and fiber-optic cables are also supported. When using coaxial cable, ARCnet
allows a maximum cable distance of about 2,000 ft from a workstation to the hub. The
maximum distance for a bus segment is only 1,000 ft. When using unshielded twistedpair cable, ARCnet supports a maximum cable distance of 800 ft.
• Topologies Star-bus
• Transmission Baseband
• Access method Token passing
• Bandwidth 2.5Mbps or 20Mbps
• Cable type(s)
RG-62 and RG-59 coaxial cable, UTP, or fiber cable
• IEEE specification None (closest to IEEE 802.4)
Server Benchmarks
Network designers and technicians are always concerned with server and network performance. By measuring performance, changes in network performance can be evaluated (i.e., determining the impact of adding new users). Performance is measured with
benchmarks—test software that can be run on servers and/or workstations. In many
cases, benchmarks are run to establish a baseline for server/network operation, and
then run again when repairs or changes are made to measure differences in operation
(and possibly identify problem areas). This part of the chapter discusses TPC, NetBench, and ServerBench benchmarks.
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 45
Chapter 1:
Network Hardware Concepts
45
TPC
The Transaction Processing Performance Council (TPC) is a well-known benchmark
organization founded in the early 1980s, and is represented by a wide range of vendors.
TPC members have cooperatively designed a series of tests to measure the power,
throughput, and performance of computer systems intended primarily for point-of-sale
(POS) applications. Here is a summary of the TPC benchmarks:
• The TPC-A benchmark was developed in November 1989, and measures the performance in update-intensive database environments for OLTP (online transaction processing) applications. TPC-A measures the number of transactions per
second a system can perform when driven from multiple terminals. In actual practice, TPC-A is rarely used today.
• The TPC-B benchmark was developed in August 1990, and is designed as a database stress test. It tests disk I/O (input/output), system and application execution
time, and transaction integrity. TPC-B measures throughput in transactions per
second. TPC-B is not used widely today.
• The TPC-C benchmark appeared in July 1992, as a successor to TPC-A, and is
designed to measure OLTP performance. TPC-C benchmarks are different than
TPC-A benchmarks because they’re considered “complex queries.” TPC-C measures multiple transaction types, more complex databases, and overall execution
structures, along with a mix of five concurrent transactions measured in transactions per minute (TPM). These tests are designed to test OLTP applications that
manage, sell, or distribute a product or service. This benchmark is used widely in
today’s computer environment.
• The TPC-D benchmark appeared in April 1995, and is designed to measure decision support systems (DSS). The TPC-D benchmark is used widely in today’s DSS
environment and measures applications that use sophisticated long-running
queries against large complex databases. There are 17 complex query tests, including two update tests. The TPC sets data volume points of 1GB, 10GBs, 30GBs,
100GBs, 300GBs, 1TB, 3TBs, and 10TBs. Each data volume point is considered a
different benchmark.
ServerBench
ServerBench 4.1 is the latest version of Ziff-Davis’ standard benchmark for measuring the
performance of servers in a true client/server environment. This differs from NetBench in
that ServerBench measures the performance of application servers, while NetBench measures the performance of file servers. The ServerBench setup places data and applications
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 46
Server+ All-in-One Certification Exam Guide
46
on the server, and uses client PCs as front ends to provide access to the applications.
ServerBench clients make requests of an application that runs on the server, and the
server’s capability to service those requests is reported in transactions per second (producing an overall score for the server). ServerBench 4.1 runs on IBM’s OS/2 Warp Server,
Microsoft’s Windows NT Server (for both Digital Alpha and x86-compatible processors),
Novell’s NetWare, Sun’s Solaris (32-bit SPARC and x86), Linux, and SCO’s OpenServer
and UnixWare 2.1. To test network file servers, use the NetBench utility instead.
NetBench
NetBench 6.0 is the official Ziff-Davis benchmark test for checking the performance of
network file servers. NetBench provides a way to measure, analyze, and predict how a file
server will handle network file I/O requests from 32-bit Windows clients. It monitors the
response of the server as multiple clients request data and reports the server’s total
throughput. The clients access the server with requests for network file operations. Each
client keeps track of how many bytes of data it moves to and from the server and how
long the process takes. This information is used by the client to calculate throughput for
that particular test mix. NetBench totals all the client throughputs together to produce
the overall throughput for a server. To test application servers, you should use the ServerBench utility instead. Version 6.0 of NetBench supports new response time measures for
NetBench clients that show how long a server takes to respond to each client’s requests.
There is also support for the 32-bit Windows client only for Windows 95, Windows 98,
or Windows NT (there is no support for 16-bit Windows, DOS, or Mac OS clients).
Avoiding Benchmark Problems
One of the most serious problems encountered with benchmarks is the integrity of
their numbers. You’ve probably heard that “statistics can lie,” and the same thing is true
of benchmarks. In order for benchmarks to provide you with reliable results, there are
some precautions that you must take:
• Note the complete system configuration When you run a benchmark and
achieve a result, be sure to note the entire system configuration (i.e., CPU, RAM,
cache, OS version, and so on). The benchmark may yield vastly different numbers
on different configurations of the same system.
• Run the same benchmark on every system Benchmarks are still software, and
the way in which benchmark code is written can impact the way it produces results
on a given computer. Often, two different versions of the same benchmark will
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 47
Chapter 1:
Network Hardware Concepts
47
yield two different results. When you use benchmarks for comparisons between
systems, be sure to use the same program and version number.
• Minimize hardware differences between hardware platforms A computer is an
assembly of many interdependent subassemblies (i.e., motherboard, drive controllers, drives, CPU, and so on), but when a benchmark is run to compare a difference between systems, that difference can be masked by other elements in the
system. For example, suppose you’re using a benchmark to test the hard drive data
transfer on two systems. Different hard drives and drive controllers will yield different results—that’s expected. However, even if you’re using identical drives and
controllers, other differences between the systems (such as BIOS versions, TSRs, OS
differences, or motherboard chipsets) can also influence different results.
• Run the benchmarks under the same load The results generated by a benchmark
do not guarantee that same level of performance under real-world applications. This
was one of the flaws of early computer benchmarking—small, tightly written benchmark code resulted in artificially high performance, but the system still performed
poorly when real applications were used. Use benchmarks that make use of (or simulate) actual programs, or otherwise simulate your true workload.
Chapter Review
Networks connect computers together in order to share files, resources, and even applications. A networked computer that provides resources is called a server. The computer
accessing those resources is referred to as a workstation or client. Server-based networks
allow resources, security, and administration to be handled from a single central location. Software is needed to support particular server features. Reliability is basically the
notion of dependable and consistent operation. Availability means that a server must
constantly be “up” and ready for immediate use, allowing a user to access the resources
he or she needs in real time. Hot swapping is the ability to pull out a failed component
and plug in a new one while the power is still on and the system is operating. Scalability allows administrators to select computers to fit the task now, and then add more
equipment as needs demand. Clustering allows more than one server to take on redundant roles in the network and improve performance. To improve server performance,
more than one processor can be used to perform additional tasks simultaneously.
Bus, star, and ring are the three major topologies used in current networks. A bus
topology connects all PCs in a single line (or “trunk”). Bus networks use terminators to
prevent signal bounce across the cabling. A star topology connects all PCs to a single
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 48
Server+ All-in-One Certification Exam Guide
48
central hub without the use of terminators, but a hub failure can disable the entire
network. The ring topology connects all PCs in a logical “loop,” and uses a token to
pass control of the network from system to system.
A bridge can act like a repeater to extend the effective length of a network cable, but
it can also divide a network to isolate excessive traffic or problem data. A router knows
the address of each segment, determines the best path for sending data, and filters
broadcast traffic to the local segment. A gateway can perform complex functions such as
translating between networks that speak different languages (using techniques such as
protocol and bandwidth conversion). The network interface card (NIC, also known as
a LAN adapter) functions as an interface between the individual computer (server or
client) and the network cabling.
Cabling (or network media) comes in many different configurations, including
unshielded twisted pair (UTP), coaxial cable, shielded twisted pair (STP), and fiberoptic (FO) cable. Bandwidth is simply the amount of data that can be handled by a cable
or device over a given time. Baseband transmission employs digital signaling to use the
entire bandwidth of the cable, while broadband transmission uses analog signaling
across a wide range of frequencies. Coaxial cables are available in thinnet and thicknet
versions. There are five categories of unshielded twisted pair (UTP) cable. IBM cabling
is separated into nine categories.
Networks break down data into small packages called packets. A packet is basically
made up of three parts: header, data, and trailer. A trailer usually contains errorchecking information called a cyclical redundancy check. Access methods regulate the
flow of traffic on the network. There are three major access methods: CSMA, token passing, and demand priority. There are several types of Ethernet: 10BaseT, 10Base2,
10Base5, 10BaseFL, 100BaseVG, and 100BaseX. Token Ring passes control from PC to
PC through the use of special packets (called tokens). ARCnet uses a token-passing
access method in a star-bus topology with data at rates approaching 2.5Mbps. Performance is measured with benchmarks—test software that can be run on servers and/or
workstations. ServerBench is the Ziff-Davis standard benchmark for measuring the performance of servers in a true client/server environment. NetBench is the Ziff-Davis
benchmark test for checking the performance of network file servers.
Questions
1. A networked computer that provides resources is called a . . .
a. Peer
b. Node
c. Server
d. Client
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 49
Chapter 1:
Network Hardware Concepts
49
2. A document that is loaded into your workstation’s memory so that you can edit
or use it locally is typically stored on a . . .
a. Database server
b. File and print server
c. Web server
d. Telnet server
3. The notion of dependable and consistent server operation is termed . . .
a. Scalability
b. Availability
c. Reliability
d. Redundancy
4. Grouping more than one server to perform the same job in the network is
called . . .
a. Clustering
b. Failover
c. Redundancy
d. Scalability
5. Which topology connects computers to each other in a straight line along a single
main cable called a trunk?
a. Line
b. Star
c. Ring
d. Bus
6. Which topology connects all PCs on the network to a central connection point
called a hub?
a. Line
b. Star
c. Ring
d. Bus
7. Which topology/architecture will shut down if the MAU fails?
a. Line
b. Star
c. Ring
d. Bus
200031_CH01/Bigelowx
4/20/01 4:26 PM
Page 50
Server+ All-in-One Certification Exam Guide
50
8. What kind of network hardware can also divide a network to isolate excessive
traffic or problem data?
a. Repeater
b. Amplifier
c. Bridge
d. Patch panel
9. RG-58 is a type of . . .
a. Coaxial cable
b. Shielded twisted pair
c. Unshielded twisted pair
d. Fiber-optic cable
10. 10BaseT is a form of . . .
a. Token ring
b. ARCnet
c. Gigabit Ethernet
d. 10Mbps Ethernet
Answers
1. c. Server
2. b. File and print server
3. c. Reliability
4. a. Clustering
5. d. Bus
6. b. Star
7. c. Ring
8. c. Bridge
9. a. Coaxial cable
10. d. 10Mbps Ethernet
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