The World of Computer Networking: A Primer

Index

Overview

In the last fifteen years, local area networks (LANs) have gonefrom bein g an experimental technology to becoming a key business toolused by companies w orldwide. A LAN is a high-speed communicationssystem designed to link computer s and other data processing devicestogether within a small geographic area such as a workgroup,department, or a single floor of a multi-story building. Sever al LANscan be interconnected within a building or campus of buildings, aswell, to extend connectivity.

LANs: Some Background

LANs have become popular because they allow users to share vitalcomputing r esources electronically, including expensive hardware suchas printers and CD RO M drives, application programs and, mostimportantly, the information the users need to do their jobs. Priorto the development of LAN technology, individual c omputers wereisolated froms each other and limited in their range of applicatio ns.By linking these individual computers over local area networks theirusefuln ess and productivity have been increased enormously. But a LANby its very natu re is a local network, confined to a fairly small areasuch as a building or eve n a single floor of a building. To realizethe full benefit of computer network ing, it is critical to link theindividual LANs into an enterprise-wide backbone network that connectsall of the companys employees and computing resour ces no matterhow geographically dispersed they may be.

Todays LA Ns and LAN internetworks are powerful, flexible,and easy to use, but they incor porate many sophisticated technologiesthat must work together flawlessly. For a LAN to really benefit anorganization it must be designed to meet the organiza tionschanging communications requirements. Building a LAN is a process ofchoosing different pieces and matching them together. This primer isdesigne d to help first-time LAN equipment buyers and users understandthe fundamentals of how LANs operate, what the different technologychoices are for building a LA N, and the ramifications of choosing oneoption over another. Also discussed is the concept of internetworkingor connecting disparate and geographically dispe rsed local areanetworks together to form an enterprise system, the differentte chnologies and products available to do so and the benefits andlimitations of e ach.

The Basics of Local Area Networking

Today local area networking is a shared accesstechnology. Thi s means that all of the devices attached to the LANshare a single communication s medium, usually a coaxial, twisted pair,or fiber optic cable. Figure 1 illustrates thisconcept: several computers are connected to a si ngle cable whichserves as the communications medium for all of them. The physi calconnection to the network is made by putting a networkinterface car d (NIC) inside the computer and connecting it tothe network cable. On ce the physical connection is in place it is upto the network software to manag e communications between stations onthe network.

 


When Station B sends a packet to another station on the LAN itpasses by all of the stations connected to that LAN. On the busnetwork i llustrated here the electrical signal representing the packettravels away from the sending station in both directions on the sharedcable. All stations will se e the packet but only the station it isaddressed to will pay attention to it.

In a shared media network, when one station wishes to send amessage to another station it uses the software in the worksta tion toput the message in an envelope. This envelope, called apacket , consists of message data surrounded by aheader and trailer that carryspecial information used by the network softw are to the destinationstation. One of the pieces of information placed in the packet headeris the address of the destination station.

The NIC then transmits the packet onto the LAN. The packet istransmitted as a stream of da ta bits represented by changes inelectrical signals. As it travels along the s hared cable all of thestations attached to it see the packet. As it goes by th e NIC in eachstation, the NIC checks the destination address in the packet head erto determine if the packet is addressed to it. When the packet passesthe st ation to which it is addressed, the NIC at that station copiesthe packet and th en takes the data out of the envelope and gives it tothe computer.

Figure 1 shows one source station sending asingle message packe t to one destination station. If the message thesource station wants to send i s too big to fit into one packet it willsend the message in a series of packets . On a shared access LAN,however, many stations all share the same cable. Sin ce eachindividual packet is small it takes very little time to travel to theen ds of the cable where the electrical signal dissipates. So after apacket carry ing a message between one pair of stations passes alongthe cable another statio n can transmit a packet to whatever station itneeds to send a message. In this way many devices can share the sameLAN medium.

Ethernet

The most widely used LAN technology in use today is Ethernet. It strikes a good balance between speed,price, ease of instal lation, and supportability. Approximately 80percent of all LAN connections ins talled use Ethernet.

The Ethernet standard is defined by the Institute o f Electrical andElectronic Engineers (IEEE) in a specification commonly known a sIEEE 802.3. The 802.3 specification covers rules forconfigu ring Ethernet LANs, the types of media that can be used, andhow the elements of the network should interact. The Ethernetprotocol provides the services calle d for in the Physical and DataLink Layers of the OSI reference model (please refer to the Stand ards andProtocols sidebar on page 12).

One element of the 802.3 spe cification states that Ethernetnetworks run at a data rate of 10 million bits p er second (10 Mbps) or100 million bits per second (100Mbps). This means that w hen a stationtransmits a packet onto the Ethernet medium it travels along that medium at 10 Mbps.

Another important element defined by the 802.3 specif ication is theaccess method to be used by stations connected to an Ethernet LAN . Itis called Carrier Sense Multiple Access with CollisionDetection ( CSMA/CD). In this method each station contendsfor access to the share d medium. It is possible for two stations totry sending packets at the same ti me which results in a collision onthe LAN. In Ethernet networks collisions are considered normal eventsand the CSMA/CD access method is designed to quickly r estore thenetwork to normal activity after a collision occurs.

Etherne t Media and Topologies.

An important part of designing and installing a LAN is selectingthe appropriate medium and topology for the environment. Ethernet networks can be configured in either a star or bustopology and installed using any of three differentmedia.

Coaxial cable was th e original LAN medium and it is used in what iscalled a bus topology (Figure 1 depicts a typicalbus topology). In this configuration the coaxial cable forms a singlebus to which all stations are attached. This topol ogy is rarely usedin new LAN installations today because it is relatively diffi cult toaccommodate adding new users or moving existing users from onelocation to another. It is also difficult to troubleshoot problems ona bus LAN unless i t is very small.

Figure 2 illustrates a star topology which is a morerobust topology. In a star topology each station is connected to acentr al wiring concentrator, or hub, by an individual length oftwisted pair cable. The cable is connected to the stations NICat one end and to a port on th e hub at the other. The hubs are placedin wiring closets centrally located in a building.



In a star topology all stations are wired to a centralwiring concentrator called a hub. Similar to a bus topology, packetssent from one station to another are repeated to all ports on the hub. This allows all stations to see each packet sent on the network, butonly the st ation a packet is addressed to pays attention toit.


As previously noted, Ethernet networks can be built using threed ifferent types of media: shielded and unshielded twisted pair,coaxial and fiber optic cables. By far the most common is twistedpair because it is associated with the more popular star topology. Itis inexpensive, and very easy to instal l, troubleshoot, and repair.Twisted pair cable comes both unshielded and shield ed.Unshielded twisted pair (UTP) cable used for LANs issimila r to telephone cable, but has somewhat more stringentspecifications regarding i ts susceptibility to outsideelectromagnetic interference (EMI) than commontelephone wire. Shielded twisted pair (STP), as i tsname implies, comes with a shielding around the cable to provide moreprotect ion against EMI.

Of the two types of twisted pair cable, UTP is by far t he mostcommonly used. The specification for running Ethernet on UTP iscalled 10BASE-T. This stands for 10 Mbps, basebandsignaling (the sig naling method used by Ethernet networks), overtwisted pair cable. Other Ethern et specifications include10BASE-5, which uses a thick coaxial cable, and10BASE-2 which uses a thin coaxial cable media.Toda y, 10BASE-5 is seldom installed in new Ethernet networks, and10BASE-2 is used o nly in very small office networks. An additionalstandard allows 10BASE-F Ether net to run on fiber optic cable

Token Ring

Another major LAN technology in use today is TokenRing. Toke n Ring rules are defined in the IEEE 802.5specification. Like Ethernet, the To ken Ring protocol providesservices at the Physical and Data Link Layers of the OSI model. TokenRing networks can be run at two different data rates, 4 Mbps o r 16Mbps.

The access method used on Token Ring networks is called token passing. Token passing is a deterministicaccess method in wh ich collisions are prevented by assuring that onlyone station can transmit at a ny given time. This is accomplished bypassing a special packet called a token from onestation to another around a ring. A station can only send a packetwhen it gets the free token. When a station gets a free token and transmits a packet it travels in one direction around the ring,passing all of the other stations along the way. As with Ethernet,the packet is usually addre ssed to a single station, and when itpasses by that station the packet is copie d. The packet continues totravel around the ring until it returns to the sendi ng station whichremoves it and sends a free token to the next station around th ering.

Token Ring Topology and Media.

Token Ring networks use what is called a ringtopology. However, it is actually imple mented in what canbest be described as a collapsed ring that l ooks likea physical star topology (see Figure 3). In Token Ring LANs eachstat ion is connected to a Token Ring wiring concentrator, called amulti-sta tion access unit (MAU), using an individualrun of twisted pair cable. Like Ethernet hubs, MAUs are located inwiring closets.


The ring topolog y as used in Token Ring networks is a collapsed ring that looks like a physical star. Each station is connected to a Token Ring wiring connector by a single twi sted pair cable with two wire pairs. One pair serves as the inbound portion of the ring (also known as the receive pair) and the other pair serves as the outbound or transmit pair.

FDDI

Fiber Distributed Data Interfac e, commonlyknown as FDDI, provides data transport at 100 Mbps, a much higher datarate than Ethernet or Token Ring. Originally FDDI networks required fiber optic cable, but today they can be run on UTP, as well. Fiberis still p referred in many FDDI networks because it can be used overmuch greater distance s than UTP cable. Like Token Ring, FDDI uses atoken passing media access metho d. It is also usually configured in acollapsed ring, or physical star, topolog y. FDDI is used primarily asa backbone, a segment of network that links severalindividual workgroup or department LANs together in a single building.It is also used to link several building LANs together in a campusenv ironment.

Structured Wiring.

Both the Ethernet star topology a nd the collapsedring topology used in Token Ring LANs are support ed by what iscalled a structured wiring architecture. Withstr uctured wiring all of the network stations are physicallystar-wired to intelligent hubs. Intelligent hubs arehubs which can be monitored and managed by network operators. Thiscombination of a star topology and intellig ent hubs maketroubleshooting and fault isolation easier and faster because each endstation is attached to the network on its own individual port, whichmeans it can be monitored easily and, if necessary, easily turned off.In addition, st ructured wiring makes adding users to the network,moving them, or making other physical changes on the network verysimple. Since both Ethernet and Token Ring networks can use twistedpair cable and can be configured in a physical star to pology, astructured wiring architecture will support either networktechnology.

Hubs: The Central Connection Point

The hub is one of the most important elements of a LAN. It is acentral connection poi nt for wiring the network (refer to Figure 4),and all stations on the LAN are l inked to each other through the hub.The term hub is generally associated with 10BASE-TEthernet networks while the term multistation a ccess unit(MAU), is used to refer to the Token Ring wiringconcentrato r. These two LAN technologies use different media accessmethods hubs and MAUs perform different media access functionsinternally, but at one level they perfo rm the same function: they areboth network wiring concentrators.


The cor nerstone of the network is the intelligent hub, orconcentrator, which serves as the control point for systems activity,management and growth. By integrating a ny combination of connectivity,internetworking and management capabilities into intelligent hubs,network managers can create the perfect physical networkinfr astructure for their environment.

A typical hub has multiple user ports to which computers andperipheral devices such as s ervers are attached. Each port supports asingle 10BASE-T twisted pair connectio n from a network station. Whenan Ethernet packet is transmitted to the hub by one station it isrepeated, or copied, over onto all of the oth er portsof the hub. In this way all of the stations see everypa cket just as they do on a bus network, so even though each stationis connected to the hub with its own dedicated twisted pair cable ahub-based network is stil l a shared media LAN picture it as aLAN in a box.

Manageable Hubs.

Intelligent hubs have been defined as manageablehubs , meaning that each of the ports on the hub can beconfigured, monitored, a nd enabled or disabled by a network operatorfrom a hub management console. Hub management can also includegathering information on a variety of network param eters, such as thenumbers of packets that pass through the hub and each of its ports,what types of packets they are, whether the packets contain errors,and h ow many collisions have occurred. Each hub vendor has some typeof management p ackage they sell with their product. These applicationsvary in how much informa tion they can gather, what commands can beissued, and how the information is pr esented to the networkoperator.

Stand-alone Hubs.

Both hubs a nd MAUs come in three configurations:stand-alone hubs, stackable hubs,and modular hubs. Some products are combinations ofthe best configurations. Stand-alone hubs are as the t ermimplies single box-level products with a number of ports.Stand-alo ne hubs usually include some method of linking them to otherstand-alone hubs either by connecting them together with alength of 10BASE-5 coaxial cabl e or cascading them using twisted pairbetween individual ports on each hub (see Figure 5). Stand-alone hubsare usually the least expensive type of hub and ar e often not managed.They are best suited for small, independent workgroups, dep artments oroffices typically with fewer than 12 users per each LAN.


Netw ork A illustrates four 10BASE-T hubs connected together bya single cable. This cable could be a coaxial or an optical fibercable. All of the hubs are part of a single LAN. Network B illustratestwo 10BASE-T hubs cascaded. Here the cable c onnecting the two ports isunshielded twisted pair wire. All of the hubs that ar e cascaded inthis fashion are part of a single LAN.

Stackable Hubs.

A third type of hub is the stackabl e hub.Stackable hubs look and act like stand-alone hubs except that se veralof them can be stacked or connected together, usually bysho rt lengths of cable. When they are linked together they act like amodular hub in that they can be managed as a single unit. Onemanageable hub, used within a stack, can typically provide themanagement for all other hubs in the stack. T hese hubs are ideal whenan organization wants to start with a minimal investmen t but knowsthat its LAN will grow. By utilizing stackable hubs, an organizatio ndoesnt need to invest in a large chassis, which may only haveone or tw o cards in it for a considerable length of time until moreare needed.

Linking Hubs.

Each hub usually represents a singl e LAN. In most organizations itis desirable to interconnect all of the LANs, w hich means linking hubsin some way. One way to link hubs is to use an i nter-repeaterlink or cascaded segment. This type of connection simply repeats all of the packets from one hub to the other hub it is linkedto, so th at in effect the two hubs are part of the same LAN.

Modular Hubs.

Modular hubs are popular in networks because theyare easil y expanded and always have a management option. A modularhub starts with a cha ssis, or card cage, with multiple card slots,each of which accepts a communicat ions card, or module. Each moduleacts like a stand-alone hub; when the communi cations modules areplaced in the card slots in the chassis, they connect to ac ommunications backplane that links them together so that a stationconnected to a port on one module can easily communicate with astation on another module. F igure 6 illustrates a modular hub.



Modular hubs provide a central point where multipleconce ntrators located in different wiring closets can be united into aLAN or WAN. Th e modular hub can be equipped with a wide variety ofconnectivity and network ma nagement modules designed to provide acustomized solution for the creation of e nterprise-wide LANs andWANs.

Modular hubs typically range i n size from four to fourteen slots,so the network can be easily expanded. Typi cally several slots in amodular hub will be filled with 10BASE-T Ethernet modul es. Forinstance, with ten modules, each supporting twelve users, a single hub could support 120 users over 10BASE-T. The modules are linked by thehigh-speed backplane, which can also be used to connect thecommunications modules to a ma nagement module that manages all of thecards in the chassis. In addition to us ing one management module fora large number of ports, all of the modules share a common powersupply. Another advantage of some modular hubs is that Ethernet, Token Ring, and even FDDI communications modules can be placed in thesame chas sis, using the same common power supplies.

Internetwork ing

The term internetworking refers to linkingind ividual local area networks together to form a single internetwork.This interne twork is sometimes called an enterprisenetwork because it inte rconnects all of the computer networksthroughout the entire enterprise. Workgr oup LANs on different floorsof a building or in separate buildings on a busines s campus can belinked together so that all of the computing systems at that sit e areinterconnected. Geographically distant company sites can also be tiedtog ether in the enterprise-wide internetwork.

An individual LAN is subject to limits on such things as how far itcan extend, how many stations can be conn ected to it, how fast datacan be transmitted between stations, and how much tra ffic it cansupport. If a company wants to go beyond those limits link more stations than that LAN can support, for example it mustinstall a nother LAN and connect the two together in an internetwork.

There are tw o main reasons for implementing multiple LANs andinternetworking them. One is to extend the geographic coverage of thenetwork beyond what a single LAN can su pport to multiplefloors in a building, to nearby buildings, and to rem ote sites. Theother key reason for creating internetworks is to share traffic loadsbetween more than one LAN. A single LAN can only support so muchtraffic. If the load increases beyond its carrying capacity userswill suffer reduced t hroughput and much of the productivity achievedby installing the LAN in the fir st place will be lost. One way tohandle heavy network traffic is to divide it between multipleinternetworked LANs.

There are three major types of dev ice used for internetworking:bridges, routers, and switches. Today the most co mmonly usedinternetworking devices are high-speed routers, especially inwide-a rea internetworks linking geographically remote sites. Butrouters are also hea vily used in building and campus internetworks, aswell. Bridges have also been popular even though the offer lessfunctionality than routers because they are less expensive topurchase, implement and maintain.

LAN switches are a n ew class of internetworking device, and manypeople believe that switched intern etworks will become the most commondesign for linking building and campus LANs in the future.Todays LAN switches and switching hubs are the first steps on amigration path to something called a synchronous transfer mode(AT M) switching, an emerging technology that will be widelyimplemented in both LANs and wide-area networks in the comingyears.

Bridges and Rout ers

Bridges and routers are bothspe cial kinds of devices used for internetworking LANs thatis, linking di fferent LANs or LAN segments together. Manyorganizations have LANs located at sites that are geographicallydistant from each other. Routers were originally designed to allowusers to connect these remote LANs across a wide-area network, butbridges can be used for this purpose, as well. By placing routers orbridg es on LANs at two distant sites and connecting them with atelecommunications li nk, a user on one of the LANs can accessresources on the other LAN as if those resources were local.

Bridges and routers link adjacent LANs. Local bri dges and routerswere first used to extend the area a network could cover by all owingusers to connect two adjacent LANs to maintain performance by reducingthe number of users per segment. Both Ethernet and Token Ring specifylimits on ma ximum distances between workstations and hubs, hubs andhubs, and a maximum numb er of stations that can be connected to asingle LAN. To provide network connec tivity for more people, orextend it to cover a larger area, it is sometimes nec essary to linktwo different LANs or LAN segments. Bridges and routers can both provide this function.

Today, however, these internetworking devices ar e also increasinglyused to segment LANs to maintain performanc e byreducing the number of users per segment. When users on a single LANbegin to experience slower response times, the culprit is oftencongestion: too much traffic on the LAN. One method users areemploying to deal with this is to brea k large LANs with many usersinto smaller LANs, each with fewer users. Adding n ew network usersmay require the organization to create new LANs to accommodate them.Implementing new applications on an existing LAN can create so muchincrem ental traffic that the organization may need to break the LANinto smaller LANs segments to maintain acceptable performancelevels.

In all of these cas es, it is still critical that users on one LANbe able to reach resources on oth er LANs within the organization. Butthe LANs must be connected in such a way t hat packets arefiltered, so that only those packets that need topass from one LAN to another are forwarded across the link. Thiskeeps the p ackets sent between two stations on any one LAN fromcrossing over onto the othe r LANs and thereby congesting them. Ageneral rule of thumb says that 80 percen t of the packets transmittedon a typical workgroup or department LAN are destin ed for stations onthat LAN. Both bridges and routers can be used to segment LA Ns.

Bridges.

Bridges are the simpler, and often less expensive , type of device.Bridges filter packets between LANs by making a simpleforward /dont forward decision on each packet they receive fromany of the networ ks to which they are connected. Filtering is donebased on the destination addr ess of the packet. If a packetsdestination is a station on the same seg ment where it originated, itis not forwarded. If it is destined for a station on another LAN, itis connected to a different bridge port and forwarded to that port.Many bridges today filter and forward packets with very little delay,mak ing them good for large traffic volumes.

Routers.

Routers are more complex internetworking devices and are alsotypically more expensive than bridges. They use Network LayerProtocol Information within ea ch packet to route it from oneLAN to another. This means that a router must be able to recognizeall of the different Network Layer Protocols that may be used on thenetworks it is linking together. This is where the termmultipr otocol router comes from a devicethat can route using many di fferent protocols. Routers communicatewith each other and share information th at allows them to determinethe best route through a complex internetwork that l inks manyLANs.

Switches

Switch es are another type of device used to linkseveral separate LANs and pr ovide packet filtering between them. ALAN switch is a device with multiple por ts, each of which can supporta single end station or an entire Ethernet or Toke n Ring LAN. With adifferent LAN connected to each of the switchs ports, it canswitch packets between LANs as needed. In effect, it acts like a veryf ast multiport bridge packets are filtered by the switch basedon the des tination address.

Switches are used to increase performance on anorgani zations network by segmenting large networks into manysmaller, less cong ested LANs, while still providing necessaryinterconnectivity between them. Swi tches increase network performanceby providing each port with dedicated bandwid th, without requiringusers to change any existing equipment, such as NICs, hubs , wiring, orany routers or bridges that are currently in place. Switches can al sosupport numerous transmissions simultaneously.

Deploying technology c alled dedicated LANs isanother advantage of using switches. E ach port on an Ethernet switchsupports a dedicated 10 Mbps Ethernet LAN. Usual ly these LANscomprise multiple stations linked to a 10BASE-T hub (see Figure 7) ,but it is also possible to connect a single high-performance station,such as a server, to a switch port. In this case, that one stationhas an uncontested 1 0 Mbps Ethernet LAN all to itself. Packetsforwarded over it from other ports o n the switch will never produceany collisions because there are no other statio ns on the LAN at thatport.


Using LAN switches a llows a network designer to create severalsmall network segments. These smaller segments mean that fewerstations are competing for bandwidth, thereby diminish ing networkcongestion.

As was noted earli er, LAN switching is a relatively new technology.Todays switching device s switch relatively large,variable-length LAN packets between different local a rea networks.Asynchronous transfer mode (ATM) is another type of switchingtech nology that switches small, fixed-length cells containing data.ATM networks can be run at much higher data rates than todaysLANs. Eventually they will be used to carry voice, video, andmultimedia traffic, as well as computer-gene rated data over both shortand long distances. ATM will be the dominant enterpr ise networkingtechnology of the future, and many companies are beginning to dev elopstrategies to incorporate ATM in their existing LANs and LANinternetworks.

Networking Today

LAN technology is evo lving. In the early 1980s LANs were strictlylocal area networks, linking small groups of computers in companydepartments. As workgroup LANs proliferated ove r the past ten years,users began connecting them to form internetworks, first w ith bridgesand later with routers. Todays networks typically comprise a combination of workgroup and campus hubs, bridges and routers.Switches are beg inning to become more prevalent, as well.

The next few years will see ne tworks evolve to include moresophisticated LAN switches and switching hubs. The y will be designedusing several different types of components, both old and new .Ethernet and Token Ring LANs will be built with stackable workgrouphubs which , in turn, will be interconnected by larger modular hubsthat may incorporate LA N switching functionality. Large networks willinclude another layer of consoli dation with networkcenter hubs linking workgroup hubs and swit ches. Routerswill continue to be used as gateways to the wide-area network lin kingother buildings and remote sites.

For networks to deliver the perfo rmance todays usersrequire, their many components must work together to deliver seamlessconnectivity between all of the users and computing systems thr oughoutthe enterprise. Flexibility to grow, power to support applications,and seamless connectivity are what users expect in the products theychoose to buil d LANs and enterprise networks.

About Bay Networks

Born from the merger of SynOptics Communications and WellfleetCommuni cations on October 20, 1994, Bay Networks, Inc., is one of theworlds lar gest networking companies with revenues exceeding $1billion and earnings of ove r $150 million.

Headquartered in Santa Clara, California, Bay Networks m anufacturesand markets a comprehensive line of networking equipment used to bui ldboth small and large-scale corporate networks for companies around theworld.

Through both direct and indirect channels, the company sells acomplete line of intelligent hubs, high-speed switches, multiprotocolrouters and sophis ticated network management systems to virtuallyevery Fortune 100 company.

The foundation of Bay Networks networking solutions is its systemof intellig ent hardware and software products. Designed to meetcurrent and future networki ng needs, these solutions provide theflexibility to create a network today that can easily grow into avast, multi-enterprise network in the future.

Ba y Networks product portfolio includes: modular,multiprotocol intelligent hubs for both network center and wiringcloset applications, highly scalable, h igh performance multiprotocolrouters for corporate and branch office connectivi ty, multi-serviceWAN switches, fixed configuration stackable work grouphubs for Token Ring, FDDI and Ethernet environments, standaloneAsynchrono us Transfer Mode (ATM) and 10/100 Mbps Ethernet switches,and a comprehensive ne twork management system that allows forsophisticated control and monitoring of these devices.

Bay Networks markets these products to large and small en d-userorganizations through a combination of original equipmentmanufacturers ( OEMs), distributors, value-added resellers and a directsales force. Typical ta rget users include worldwide retailers, foodservice companies, financial instit utions, technology manufacturers,telecommunications companies, hospitals and un iversities, andgovernment organizations.

A representative list of Bay N etworks customers include: AT&T,Australia Department of Social Services, Bank o f InternationalSettlements, Bear Stearns, Boeing Aircraft, British Petroleum, C haseManhattan Bank, Ford Motor Company, General Motors, McDonalds,MCI, Northwestern Mutual Life, Sprint, 3M, and Wal-Mart.

The company has a co mbined market value in excess of $2.5 billionand employs a staff of over 3,000 around the world. As a market shareleader with over 23 percent of the estimate d $4.6 billioninternetworking market, Bay Networks has an installed base of ove r 10million end-user connection for its intelligent hubs products and morethan 200,000 port connections installed for its router products.

The company pioneered the networking industry in the mid-eightiesby innovating the ability to run Ethernet networks over common phonewire as well as being one of the fir st companies to bring to markethigh-speed, multiprotocol routing.

Addit ionally, Bay Networks has a number of strategic developmentand technology partn erships with a variety of industry leadingcompanies including: IBM, Novell, Int el, Hewlett-Packard and SunMicrosystems.

Publicly held and traded on NA SDAQ, Bay Networks is led by PaulSeverino (Wellfleet), chairman of the board an d by Andy Ludwick(SynOptics) president and CEO.

Glo ssary of Terms

Asynchronous Transfer Mode (ATM) A typeof switching technology in which the switches are small, fixed-len gthcells containing data.

Backbone A segment of network that linksseveral individual workgroup or department LANs together i n a singlebuilding. It is also used to link several building LANs together in acampus environment.

Bridges Bridges filter p ackets betweenLANs by making a simple forward/dont forward decision on e achpacket they receive from any of the networks to which they areconnected.

Carrier Sense Multiple Access with Collision Detection(CSMA/CD) An element defined by the 802.3specification. It is an access method which is used by stationsconnected to an Ethernet LAN. In this method e ach station contendsfor access to the shared medium.

Collision When two stations try to sendpackets at the same time. In Eth ernet networks collisions areconsidered normal events and the CSMA/CD access me thod is designed toquickly restore the network to normal activity after a colli sionoccurs.

EMI Electromagnetic interference.

Ethernet The most popular LAN technologyin u se today.

Fiber Distributed Data Interface (FDDI) LAN technology that runs at 100 Mbps, a much higher data rate thanEthernet or Token Ring. Originally FDDI networks required fiber opticcable, but today t hey can be run on UTP, as well.

IEEE 802.3 An Ethernet specificationcommonly defined by the Institute of Electrical and Elect ronicEngineers (IEEE). The 802.3 specification covers rules for configuringEth ernet LANs, the types of media that can be used, and how theelements of the net work should interact.

IEEE 802.5 A Token Ring specificationcommonly defined by the Institute of Electrical and ElectronicEng ineers (IE

Intelligent Hubs Intelligent hubs a rewiring concentrators which can be monitored and managed by networkoperators.

Inter-repeater link One method of linkinghub s. This type of connection simply repeats all of the packets fromone hub to the other hub it is linked to, so that in effect the twohubs are part of the same LAN.

LAN Internetwork Connecting disparate and geographically dispersed local area networks together to form anenterprise sys tem.

Local Area Network (LAN) A LAN is ahigh- speed communications system designed to link computers and otherdata processing devices together within a small geographic area suchas a workgroup, department , or a single floor of a multi-storybuilding.

Manageable Hubs Another definition forintelligent hubs. Each of the ports on t he managed hub can beconfigured, monitored, and enabled or disabled by a networ k operatorfrom a hub management console.

Modular hubs A modular hub starts with achassis, or card cage, with multiple card s lots, each of which canaccept a communications card, or module. Each module ac ts like astand-alone hub; when the communications modules are placed in thecar d slots in the chassis, they connect to a high-speed communicationsbackplane th at links them together so that a station connected to aport on one module can e asily communicate with a station on anothermodule.

Network Inte rface Card (NIC) The physicalconnection from the computer to the network is made by putting anetwork interface card (NIC) inside the compute r and connecting it tothe shared cable.

Multi-station Access Un it (MAU) A TokenRing wiring concentrator which connects each station in a Token RingLAN.

Open Systems Interconnect Reference Model (OSI) A communications model developed by the Internati onalStandards Organization (ISO) to define all of the services a LANshould pro vide. This model defines seven layers, each of whichprovides a subset of all o f the LAN services. This layered approachallows small groups of related servic es to be implemented in a modularfashion that makes designing network software much more flexible.

Packet In a shared media n etwork, whenone stations wishes to send a message to another station, it uses t henetwork software to put the message in an envelope. This envelope iscalled a packet.

Routers Routers are more complexint ernetworking devices and are also typically more expensive thanbridges. They u se Network Layer Protocol Information within eachpacket to route it from one LA N to another.

Shared Access Shared media techn ologymeans that all of the devices attached to the LAN share a singlecommunica tions medium, usually a coaxial, twisted pair, or fiber opticcable.

Shielded twisted pair (STP) Cable whichhas shielding aro und it to provide more protection againstelectromagnetic interference (EMI).

Stackable hubs Stackable hubs look and act like stand-alone hubs except that several of them can be stacked orcon nected together, usually by short lengths of cable. When they arelinked togeth er they can be managed as a single unit.

Stand-alone hubs Single box-level hubswith a number of ports. Stand-alone hubs usua lly include some methodof linking them to other stand-alone hubs eithe r by connecting themtogether with a length of 10BASE-5 coaxial cable or cascadi ng themusing twisted pair between individual ports on each hub.

Structured wiring architecture A structuredwiring architectu re which physically star-wires all network stationsto intelligent hubs.

Switches Links several separate LANs andprovides pa cket filtering between them. A LAN switch is a device withmultiple ports, each of which can support an entire Ethernet or TokenRing LAN.

10BA SE-T The specification for runningEthernet on UTP. This stand s for 10 Mbps, baseband signaling (thesignaling method used by Ethernet network s), over twisted paircable.

10BASE-5 An Ether net specifications whichuses a thick coaxial cable. 10BASE-5 is seldom install ed in newEthernet networks today.

10BASE-2 An Ethernet specification thatuses a thin coaxial cable medium. 10BASE-2 is only used in very smalloffice networks.

Token Passing The access method used onToken Ring networks.

Token Ring A major LAN technology in usetoday. Token Ring rules are define d in the IEEE 802.5 specification.Like Ethernet, the Token Ring protocol provid es services at thePhysical and Data Link Layers of the OSI model. Token Ring n etworkscan be run at two different data rates, 4 Mbps or 16 Mbps.

Unshielded twisted pair (UTP) UTP cableis similar to telep hone cable, but has somewhat more stringentspecifications regarding its suscept ibility to outside EMI than commontelephone wire. UTP is used much more often than STP.

Standards and Protocols

LANs ar e complex systems that implement many different services inorder to provide com munication between all of the types of devicesthat can be connected to them. A communications model called the OpenSystems Interconnect (OSI) Reference Model was developed by theInternational Standards Organization (ISO) to define all o f theservices a LAN should provide (see Figure 8). This model definesseven la yers, each of which provides a subset of all of the LANservices. This layered approach allows small groups of relatedservices to be implemented in a modular fashion that makes designingnetwork software much more flexible. A network sof tware module thatimplements services at the Network and Transport Layers of the modelcan be paired up with different Physical and Data Link Layer modulesdepe nding on the requirements of the users application.


The Internation al Standards Organization (ISO), the primarystandard-setting body in the data c ommunications industry, developed aframework for LAN standards called the Open Systems Interconnectionreference model. This reference model represents a stand ard approachto communicate information throughout a network so that a variety o findependently developed computer and communications devices caninteroperate.< /strong>

But the OSI model doesnt say how th ese services should actually beimplemented in LAN equipment. The how to part has been defined in anumber of different protocols that have been developed byinternational standards bodies, individual LAN equ ipment vendors, andad hoc groups of interested parties. These protocols typica lly definehow to implement a group of services in one or two layers of the OSI model. For example, Ethernet and Token Ring are both protocols thatdefine diff erent ways to provide the services called for in thePhysical and Data Link Laye rs of the OSI model. They have both beenapproved by the Institute of Electrica l and Electronic Engineers(IEEE), an international communications standards body.

Because they are approved and published by the IEEE, both theEthernet and Token Ring protocols are said to be industry standards.Any company can acquire the specificat ions and design Ethernet orToken ring NICs. Users can purchase an Ethernet NIC, for example,from any vendor and be assured that it will operate in a network withEthernet NICs from other vendors. This degree ofinteroperability is highly desirable.

However, there are many more protocols for providing services atthe higher layers of the OSI model and very few of them have beenapproved by an international standards bodies. In fact, most upperlayer LAN protocols are incorporated into proprietary networkoperating systems, such as Novells NetWare, IBMs LAN Server, andMicrosofts LAN Manager. A user has to buy only that vendorsproducts in order to be assured that they will interoperate on aLAN.

Network Operating Systems

Ethernet and Token Ring technologies are just one part of acomplete LAN. They provide the services specified in the Physical andData Link Layers of the OSI model, but several other services must beadded on top o f the connectivity of Ethernet or Token Ring. Networkoperating systems (NOSs) are most often used to provide the additionalcommunications services.

A NOS defines client and server systems. Clients are individualuser workstations attached to the ne twork where application programsare run and data is generated. Servers are shared network resourcesthat provide hard disk space for users to store files, printerservices, and a number of other network services. The networkoperating system provides a set of protocols in software that run onboth servers and client systems and allow them to communicate witheach other, share files, printers and other network resources.