Table of Contents   Chapter 1   Chapter 3   Netguru.net   Glossary

Chapter 2 - Local Area Networks

This chapter will introduce you to networking concepts, terminology, and technology from the perspective of the local area network. Since most networking personnel get their feet wet in local area networks (LANs) as opposed to larger wide area networks, this seems the appropriate place to start. This chapter will approach the technology of networking by migrating from a general view to one of more specifics in order to fully cover the topic. Perhaps the best place to begin is to look at a definition of a local area network.

Local Area Network - An interconnection of computers and peripheral devices contained within a limited geographical area utilizing a communication link and operating under some form of standard control.

The interconnection mentioned above follows a physical and logical layout. This layout, called a topology, governs many aspects of LANs including how they function and how easy they are to troubleshoot.

Point-to-point topology is the simplest of the physical layouts of network devices. Point-to-point connections mean that two devices (nodes) have a single path for data to travel between them and there is nothing that breaks up that path.

Figure 2-1: Point-to-Point connections can be established between many devices.

A prime example of how this topology is implemented in networking is the manner in which terminals are now connected to mainframes or mini-computers. Instead of having many cables from numerous terminals hooked into one of these computers, a device known as a terminal server allows the data from several terminals to be transmitted over a single cable. This single cable connection between the computer's front-end processor and the terminal server forms a point-to-point link. In addition, some terminal servers form point-to-point links with the individual terminals (Figure 2-1).

The point-to-point topology can be seen as one of the basic building blocks of larger, more complicated topologies. All major topologies include point-to-point connections, even if there is no wire between two devices, but some other medium instead. Satellite transmissions are considered to be point-to-point communications. Similarly, laser transmissions can also be viewed in this manner. A variant on point-to-point connections is a multipoint topology in which a single cable may split into several segments in order to connect to several devices.

Point-to-point topology is not just limited to networking use. You should be aware that the direct connection of a PC to a printer follows a point-to-point topology. In fact, any externally connected device, including modems or hard disk drives would also fall under this classification.

If you have ever had the occasion to visit San Francisco, you might have noticed that the world-famous streetcars in that scenic city utilize a common cable running beneath the streets to propel them up the steep hills. Similarly, other major cities have mass transit systems like busses that utilize common wires above the streets for power. These shared cables might be called "bus wires", an excellent description of one of the most popular topologies for LANs -- the bus topology.

Just as in the example of the electric busses, all devices share a common wire to transmit and receive data through in the bus topology (Figure 2-2). This approach is very economical as a single cable is cheaper to purchase than several individual cables for each device. Additionally, a single cable is easier to install than several cables. These apparent advantages of the bus topology are offset, however, by the difficulty in troubleshooting a problem in this layout scheme. Since all devices use the common wire, how do you track down one that has gone bad? The worst-case scenario in this situation may involve a trouble-shooter visiting and detaching every unit on the LAN until the offending machine is located. In short, troubleshooting bus topologies may require a good pair of sneakers.

On the ends of the common cable or bus, a device called a terminator is utilized to absorb signals that have traversed the entire length of the bus. Since everyone shares the same cable, no two machines can transmit at once or the bits of data from each will collide destroying both pieces of information. This event is called a collision and obviously too many of them can be disastrous to traffic flow on a network. The terminator's vital role in absorbing data that has traveled the whole network ensures that bits of information do not reflect back across the bus. A data reflection can occur any time an electronic signal encounters a short (where the wires inside a cable get connected accidentally) or an open (where the wires are left unconnected as in when a terminator is removed). The end result is the same -- reflected data collides with the "good" data on the LAN and traffic flow is impacted.

Figure 2-2: Devices all share a common cable for transferring data in a bus topology LAN. Signals are eventually absorbed by the terminator.

Ethernet is perhaps the most common type of LAN utilizing bus topology. Ethernet's speed is quick (10 Mbps) provided there are not a lot of collisions occurring. Ethernet handles collisions by allowing machines that experience this event to retransmit their data again at different timed intervals. Ethernet also incorporates a mechanism to warn all devices when collisions have occurred so they will not interfere with the error-correcting process. More details on this technology are provided in a later section of this book.

Ethernet is not the only type of network that can operate using a bus topology. ARCnet, briefly mentioned in Chapter 1, can also function in a bus type of scheme. ARCnet differs from Ethernet in that every device must take its orderly turn to transmit data. Consequently, collisions aren't a problem for ARCnet. One minor problem that often catches administrative personnel off guard is the accidental use of an Ethernet terminator on an ARCnet LAN. If unlabled, which is very common, the terminator could inadvertently be installed on the ARCnet network severely impacting, if not halting, network performance. Terminators differ in their resistance value (in ohms), which can be measured with any multimeter device. Ethernet's terminators should measure at 50 ohms while ARCnet's should measure at 93 ohms.

Token Ring technology allows for devices with token ring cards in them to revert to a bus topology in the event their usual topology (a ring) is broken. For token ring systems, a small signal called a token is passed from one machine to another in a certain order. This signal gives permission for the device to transmit data. Unlike Ethernet bus systems, absorption of data that is already "used" is done by the token ring network interface cards (NICs) themselves, not by an external termi-nator. Though tokens are mostly used with ring topologies (to be discussed later), standards exist for what is known as a token bus. This specification details how a token is used in a bus topology.

In summary, the key strength for the bus topology is its minimal use of cable and ease of installation. Its major drawback is its difficulty to troubleshoot.

Today if you decide to install a LAN, your local LAN dealer will probably suggest you look seriously at star topology networks. Star topology networks are nothing new, they just offer some benefits that are hard to overlook. Star topology derives its name from the arrangement of devices so that they radiate from a central point. At the central point we usually see a device generically called a hub (Figure 2-3).

Key to the benefits of the star topology is the hub unit which may vary in function from a simple signal splitter (called a passive hub) to one that amplifies and keeps statistics on data traveling through them (termed as an active and intelligent hub). In fact, hubs may be sophisticated enough to selectively disconnect any machine connected to them that is misbehaving, as well as allow network operators to dial into to them and monitor the performance of a single workstation. It's these advantages that make the star topology a popular choice in the networking marketplace. Hubs that amplify signals coming through are called active hubs or multiport repeaters.

Star topologies do require more cable than a simple bus topology, but most use a relatively inexpensive type of cable called twisted pair cabling which helps control costs of wiring. The hubs themselves require expense and the level of that expense is directly attributable to how complex a hub is needed.

Troubleshooting a star topology network is a bit easier than bus topology. At the very least, one may disconnect devices from a central hub to isolate a problem as opposed to visiting each individual machine. Above this physical level of troubleshooting, there is hub management software that can report problems back to you. It's obvious how the central hub device offers advantages, but there is one drawback. The hub itself represents a single point of failure. If you lose a hub, you effectively lose all workstations attached to it. Quality and reliability of hub products you purchase can not be over-stressed.

Figure 2-3: The star topology involves one or more devices radiating out from a central point (i.e. hub).

Ethernet, ARCnet, token ring, and FDDI (fiber optic) LANs all use the star topology in some form of their implementation. The flavor of Ethernet that uses this layout is called 10BASE-T Ethernet. In this system, the hubs are referred to as concentrators. Each concentrator amplifies each signal passing through it so that data will travel further along the cabling connecting each workstation to the device. Each piece of cable radiating out from the hub device, along with its workstation, is known as a lobe.

ARCnet has used the star topology for some time. Hubs for ARCnet comes in active and passive varieties. Passive hubs do not perform any re-amplification of signals passing through them. For that reason cables stretching out from passive hubs seldom extend beyond 100 feet in length. However, for about $50.00 more, one may purchase an active hub that does amplify signals allowing lobes to extend up to 2,000 feet from the device. That's a marked improvement and, in the opinion of this author, worth the extra 50 bucks.

Token ring systems are unique in that physically they use a star topology, but logically they use what is known as a ring topology. The advantages of star topology are apparent in this type of LAN with its central hub device being referred to as a Multistation Access Unit (MAU, MSAU) or Controlled Access Unit (CAU). Inside this central device a ring is created connecting all lobes. This ring is what the token uses to travel from machine to machine on the network.

The specification for networks utilizing fiber optic cable is called FDDI, short for Fiber Distributed Data Interface. FDDI is often pronounced like "fiddy". FDDI networks are very similar to token ring networks in their layout and operation. Therefore, they share many of the same advantages, including a star topology. FDDI networks operate at speeds of up to 100 Mbps.

There are a few other points worthy of noting concerning the star topology. It is also used by the AT&T-marketed StarLAN. StarLAN shares many characteristics with 10BASE-T Ethernet. Stars can require a lot of cable depending on how they must be physically laid out. Incidentally, the earliest commercially form of a Novell network operating system (NOS) used a star topology where every workstation had to plug into a single proprietary file server. Sold in the early 80s, this system supported a maximum of 24 workstations. Speeds on the network were rated at 232,000 bits per second (232 Kbps). Fortunately, Novell has come a long way since then, but, then again, so has everybody.

Star topologies are not just used by local area networks. They are found in other areas of life as well. One of the most common star topologies found in any company is the phone system, which typically radiates out from a central site. Because of this, phone system equipment and cabling are often used in setting up LANs.

In summary, star topology systems offer better troubleshooting and management capabilities, but require more physical resources than a comparable bus system.

Ring Topology describes the logical layout of token ring and FDDI networks. In this scheme, a ring is created to which each device (workstation, server, etc.) attaches (Figure 2-4). A special signal, called a token travels around this ring visiting each machine, letting it know that it is that machine's turn to transmit. Since the token visits every node, every one gets the chance to transmit, creating a very "fair" LAN. This simplistic explanation belies the true complexity of ring topology systems available today. Token ring LANs, and their FDDI cousins, are the most sophisticated, fault-tolerant, and, consequently, the most ex-pensive systems available in the current marketplace.

The logical creation of a ring allows information on such a LAN to travel in one direction. Since only one device is allowed to transmit at a time, collisions are not a problem on ring systems. Of course there are always problems that can occur like bad network cards or hub units that will bring a ring topology LAN to a grinding halt, but they are often very resilient. Typical ring system network interface cards (NICs) contain the ability to perform what is known as signal regeneration. This means information received by them is copied and retransmitted at a higher amplification. Since every piece of data traveling around a ring must visit each device, the signal gets regenerated numerous times. This feature allows for greater distances between nodes and increased chances that good data will completely traverse the ring. More details on ring topology systems will be passed along in later sections of this coursebook.

Figure 2-4: Even though token ring LANs utilize a star topology physically, this illustration shows that a logical ring is created inside the MAU.

Mesh topology is uncommon today because of its sheer impracticality. In a mesh topology system, every node is connected to every other node. The pervading thought behind this is to offer the maximum amount of reliability for data transit and fault-tolerance (Figure 2-5).

The major problem is the amount of cabling necessary to create this topology, plus each link from one device to another requires an individual NIC. Not only are physical components wasted, but the overall capacity to carry data is grossly under-utilized unless all nodes are transmitting to one another almost constantly.

Special topologies refer to those networks that are made of several different topologies. Another name given special topologies is hybrid topology. Special topologies are becoming quite common today as corporations continue to link their internal LANs together while adding external networks to the mix via wide area networks (WANs). This topology description usually refers to a collection of networks.

An example of a special or hybrid topology would be one where a perhaps a ring topology network (token ring) is used to connect a series of star topology LANs (10BASE-T Ethernet). Then the ring network is connected to another network via point-to-point topology (Figure 2-6).

Figure 2-6: Special or hybrid topologies are often the combination of several different kinds of other topologies.

From a troubleshooting point of view, techniques usually involve isolating sections of the whole hybrid network as to determine the source of problems. Once the troubled piece is located, techniques to finish off the problem vary by what type of topology may be used by that offending network.

These special topology networks require special management tools that are capable of interacting with several different types of LAN environments. Novell offers a product called Network Management Services that is designed to handle these systems. In addition, several companies have embraced a standard called DME (Distributed Management Environment) in creating tools for hybrid networks. DME was created by an organization called the Open Systems Foundation (OSF).

A local area network can be composed of several components. This section deals with what those components are focusing on terminology and functionality.

Server is a generic term applied to any machine running a "service" application. That service being performed might include access to shared files (file server) or access to shared printers (print server).

Novell's file services are all governed by the portion of the Novell oper-ating system that resides on your file server. In addition, NetWare provides security services that offer login/password protection.

Figure 2-7: Several different types of servers are utilized on LANs.

There are other types of servers besides file and print servers. Communication servers offer access to remote devices outside of a network. That access might be to a mainframe or minicomputer, or other networks, workstations or servers. Typically, a machine that allows multiple users to share one or more modems for external connections is called a modem server. Modem servers are becoming increasingly popular today as more and more companies find the need to access external information or E-mail services.

Another type of server is known as a database server. This unique device assists users in interacting with databases by coordinating the data sent to the local workstation. It takes a burden off the local PC by filtering out all but required data, which also greatly reduces LAN traffic.

File servers sit at the heart of just about every network. Their responsibility is to dole out files to users requesting them and to sometimes deny that access where appropriate. File servers must know which directories and files that certain users are allowed to utilize in order to efficiently manage them. The responsibility of providing security information to the machine is that of the supervisor, administrator, or some other level of network management personnel.

When users request a file, its contents are copied across the network into the memory of the user's local workstation. Once there, the user may use it however they wish. Some files are not designed to be simultaneously shared on the network. Many executable files, for instance, are only utilized by one person at a time. Consequently, if one user attempts to use one of these non-shareable files while another has it tied up, the file server will be responsible for letting the user know there is a conflict. For those files that are shareable, the file server will allow multiple copies of these to be sent to the workstations if the users only want to view the contents of them. If users are allowed to simultaneously update a file, its records being updated would have to be locked so more than one user can not be updating the same section of the file. This would pose a serious conflict and might result in the "deadly embrace". The file server must be able to distinguish whether or not a file is shareable or non-shareable. Often that delineation is done by the network administrator.

A print server's role is very important in the shared peripheral environment as it carries out the crucial task of making sure data from an application successfully reaches its temporary holding tank (queue) and subsequently the printer for which it was destined.

The queue mentioned above is a simple directory located beneath the SYSTEM directory on a NetWare file server. It is a holding place for files containing data to be printed. The concept of placing these files (called print jobs in NetWare) into a queue is called spooling, a term borrowed from mainframe vernacular.

Spooling itself must be performed in any shared printer environment in order to eliminate the possibility of two print jobs reaching the printer at the same time resulting in conflict. NetWare makes sure that only one print job gets printed on a single printer at once.

The print services may be controlled by software or hardware. Novell print services are software controlled. If you were to buy a device that attaches directly to your LAN and to your printer(s), that unit would be called a hardware print server. It carries out the role of making sure print data gets printed correctly. Examples of hardware print servers would include Intel's NetPort and Castelle's JetPress card for HP LaserJet printers. Other vendors of software-based print servers include LAN Systems with their LAN Spool product and Brightwork Development with their PS-Print product.

We should be careful to delineate that the term "workstation" may be a little misleading depending on your particular involvement in the computer industry. In PC-based local area networking, a workstation refers to a machine that will allow users access to a LAN and its resources while providing intelligence on-board allowing local execution of applications. This would pretty well cover the gamut of all PCs.

The term is also applied to some CAD (Computer-Aided Design) or CAM (Computer-Aided Manufacturing) machines that may not be attached to a LAN. In addition, there is a machine that is manufactured by Sun Microsystems, Inc., that is also called a workstation - a Sun Workstation. This device uses the UNIX operating system and can also function as a file server. Its speed and capabilities are generally superior to that of everyday PCs.

Workstations may allow data to be stored locally or remotely on a file server. Obviously, diskless workstations require all data to be stored remotely including that data necessary for the diskless machine to boot up. Executable files may reside locally or remotely as well, meaning a workstation can run its own programs or those copied off the LAN. Though the source of data doesn't matter, the destination for execution does. Processing is done on local machines in PC LANs.

The NIC is obviously a crucial component to networking. It allows a device to participate on the network. Token ring LANs require token ring NICs, Ethernet LANs require Ethernet NICs, etc.

Software is required to interface between a particular NIC and an operating system (i.e. NetWare). This interface is called a driver. NetWare provides several drivers for different vendors' cards. The vendors themselves will provide drivers for their cards as well. Different drivers are needed for integrating a NIC on a workstation as opposed to a file server. That's because the operating systems on the two types of machines are different.

4. Hubs

Hubs are a crucial element to all star topology LANs. Hubs serve as a central device through which data bound for a workstation travels. The data may be distributed, amplified, regenerated, screened or cut off.

Hubs have different names depending on the type of LAN. In token ring LANs they are referred to as Multistation Access Units or Controlled Access Units (MAUs or CAUs). In 10BASE-T Ethernet, they are referred to as concentrators. In ARCnet they are simply called hubs.

Hubs vary in their capabilities and sophistication. ARCnet passive hubs are very inexpensive and only split signals among several devices. Other hub units cost several thousands of dollars providing state-of-the-art remote management and diagnostic capabilities.

Peripherals include any device that would ordinarily be attached to a computer. LANs allow many of these devices to be shared among several workstations.

File servers, in effect, allow the sharing of one common peripheral - the disk drive. Other peripheral devices commonly shared include scanners, modems, plotters and printers.

There are two main types of software utilized in the networking environment - operating systems and applications. In the NetWare environment there are at least two operating systems utilized. NetWare is the operating system (OS) residing on and operating the file server. DOS, OS/2, Unix, or System 7 may the name of the OS on your workstation.

The NetWare operating system allows rapid access to the shared hard disk(s) of a file server. It accomplishes this feat through intelligent placement of items in the file servers on-board memory (caching) relieving the need for continual disk access. NetWare also provides security to the files as well as provides output to you that closely or exactly emulates your local OS. The local OS might include one that you directly interact with such as DOS or one that operates behind a graphical user's interface (GUI) like the Macintosh System 7 OS.

The local OS must still oversee the execution of programs locally as well as handle all requests that are to be carried out on a local basis. For instance, utilizing DOS's DIR command requires the local DOS to cough up the file listing on the requested drive. An NDIR command in NetWare demands a similar response of the file server's operating system. Part of the functionality of NetWare files residing and operating on the local workstation is to determine whether or not commands should be routed to the local operating system or NetWare.

Figure 2-8: NetWare resides on the File Server and in the shell files running on the local workstation which may itself be utilizing another operating system.

Applications for the LAN vary incredibly in their use and design. Applications that make use of certain network features (such as network drive letters) are considered to be "LAN-Aware". Some applications may be LAN-aware but limited in their functionality when in the network environment.

Communication schemes are those methods used by various types of local area networks for transferring data from one point to another. Another common term applied to this function is channel access method.

There are several schemes or methods used in networking today. The leaders include contention and token passing. Another method that has been used in the past is called polling.

The contention channel access method involves multiple devices sharing a common transmission media. An example would be Ethernet's modus operandi. In bus topology systems like this, all devices are attached to a common wire. As mentioned in a previous section of this coursebook, this means that only one device may use the common wire at a time. Since several devices may need to use the wire at once, machines are said to be contending for the media. If the system is operating within tolerable limits, every machine will eventually get the opportunity to transfer data.

Figure 2-9: With contention systems, devices must listen for the opportunity to transmit data.

Ethernet systems use a channel access method known as CSMA/CD, short for Carrier Sense Multiple Access / Collision Detection. Though this seems a lot of words, the meaning is quite simple. Carrier Sense means that each device checks the LAN before it starts transmitting to see if some other device is using the media then. If another signal (containing a "carrier") was present, then the device attempting to send would wait until the LAN is clear. Then it transmits its data. The collision detection part means that each workstation listens to make sure that only one signal is present on the LAN. In the event there are two then obviously the data from one device has collided with that of another. Once a workstation detects a collision, it sends out a series of 1 bits alerting the rest of the network. At that point everyone stops transmitting and each workstation waits a random amount of time before attempting to transmit again. The delay time is regulated by a random number generator on-board each Ethernet card.

LocalTalk LANs used by Macintosh PCs also use CSMA contention schemes, but these machines incorporate a technology called time- division multiplexing to allow avoidance of collisions. In fact, LocalTalk systems are said to be CSMA/CA systems, with CA standing for Collision Avoidance.

The major advantage of contention systems is that devices may transmit whenever they like just as long as the LAN is free. Consequently the overhead of devices waiting on the opportunity is generally low. Since any device can participate at almost any time, no attempt is made to prioritize LAN access in any way.

However, as traffic increases in a contention system, collisions can become excessive, severely impacting the overall performance of the network. The capacity of the LAN may be far underutilized in this event. The other major disadvantage is that contention systems do not follow an easily predictable pattern of performance degradation as traffic increases. The true loss in performance can only be guessed at statistically.

The Institute of Electrical and Electronic Engineers (IEEE) has created a standard for Ethernet-type systems that include specifications for implementation of contention in these types of LANs. The standard is called 802.3.

2. Token Passing Scheme

This technology is used for token ring systems. Its incorporation along with complimentary fault-tolerance capabilities yield a LAN with a fair amount of sophistication, manageability and reliability.

Figure 2-10: The token visits every device on the LAN giving each permission to transmit if ready.

In this channel access method, a small signal called a token regularly visits each device. The token gives permission for the device to transmit if it needs to. If a transfer of data is needed, the device receives a set amount of time to broadcast its data. When it is done, the machine then retransmits the token to another machine giving that recipient permission to transmit, and so the system continues. This mechanism ensures opportunity for all devices to gain access to the LAN. Because of its predictable behavior, token scheme LANs offer the advantage of priorities, where a certain group of devices may have enhanced access to the LAN if warranted.

Token passing systems may be implemented using either bus or ring topology. The IEEE standard governing token bus systems is called 802.4. The token ring specifications are called 802.5. Vendors count on standards such as these to help make sure that their products are interoperable with those of other vendors.

As traffic demand increases on a token LAN, the overall throughput of data rises as well as until a point is reached where the networks simply cannot accommodate anymore. The function in this case is somewhat like a waterwheel. The wheel itself receives water from a sluice. You may increase the capacity of the wheel, but the sluice can only hold so much water, so there is a finite limit to the throughput of the system.

Because the throughput characteristics of token LANs are so predictable, and because of the characteristics of traffic demand vs. throughput, these systems are ideal for heavy traffic situations. However, the complexity of such a LAN does come at some cost. Token systems require overhead to carry out their many functions including fault-tolerance. Plus, token ring systems are considerably more expensive than Ethernet systems. Factors weighing in deciding which system to choose should include traffic demand and budgetary restraints.

Polling is a means by which a central controlling device may regulate the opportunity for machines to transfer data on the LAN. In effect, several devices attached to a controller unit are individually given permission to access the LAN. This technology is often employed in LANs associated with mainframes and minicomputers.

The terminology used in polling systems is worthy of mentioning. The device that governs the access of other units is called a "controller" or a "primary" device. The units themselves can be referred to as "secon-daries".

When the system is operating, the controller gives permission to a secondary to send via a signal akin to a token in functionality. The secondary then has exclusive use of the network to transfer data for a set period of time. When that time expires or if the device is through transmitting, the controller routes the same request for transmit to the next secondary in line.

The obvious advantage of a polled system is the fairness factor. All devices are given access at a predetermined time. In fact, certain devices may be given more frequent access courtesy of priorities if warranted. In addition, polling systems are highly predictable in their behavior. As traffic load increases so does throughput until a certain point is reached. The example of the waterwheel as cited in the token-passing section applies to this LAN system as well.

Figure 2-11: Polling can be used by a controlling device to allow attached machines to transmit at predetermined intervals.

Like the token passing systems, polling may not provide the best use of all the LAN's capacity at lower traffic levels. Obviously there is overhead involved in sending out requests to transmit, especially to devices that need to transmit nothing. They still have to be polled for this to be ascertained.

Transmission media is what actually carries a signal from one point to another. This may include copper wiring in the case of twisted pair cable or coax cable, or electronic waves in the case of microwave or satellite transmission. A medium such as copper wiring is referred to as bounded media because it holds electronic signals. Fiber optic cable is said to be bounded media as well because it holds light waves. Other media that do not physically constrain signals are considered to be unbounded media.

Twisted pair cabling is the current popular favorite for new LAN installations. The marketplace popularity is primarily due to twisted pair's (TP's) low cost in proportion to its functionality. Its usage has been justified through years of implementation by phone companies as it is the medium used by them to connect our world together. In many cases, TP cabling has already been installed in a site by the phone company during telephone installation removing the need to put in any new cabling for a local area network.

The construction of TP is simple. Two insulated wires are twisted around one another a set number of times within one foot of distance. If properly manufactured, the twists themselves fall in no consistent pattern. This is to help offset electrical disturbances which can affect TP cable such as radio frequency interference (RFI) and electromagnetic interference (EMI). These "pairs" of wires are then bundled together and coated to form a cable.

Figure 2-12: Twisted pair cabling is exactly what its name implies - two wires twisted around one another.

Twisted pair comes in two different varieties - shielded and unshielded. Shielded twisted pair (STP) is often implemented with LocalTalk by Apple and by IBM's token ring systems. STP is simply TP cabling with a foil or mesh wrap inside the outer coating. This special layer is designed to help offset interference problems. The shielding has to be properly grounded, however, or it may cause serious problems for the LAN. Twisted pair cabling with no shielding is simply called unshielded twisted pair (UTP).

Connectors used with TP included RJ-11 and RJ-45 modular connectors in current use by phone companies. Occasionally other special connectors, such as IBM's Data Connector, are used. RJ-11 connectors accommodate 4 wires or 2 twisted pairs, while RJ-45 houses 8 wires or 4 twisted pairs.

TP cabling has been around a while and is a tried and true medium. It hasn't been able to support high speed data transmissions until relatively recently however. New development is focusing on achieving 100 Mbps throughput on UTP without costing the user an arm and a leg. A copper version of fiber optic's FDDI, called CDDI, will continue to mature while standardization is worked out for 100 Mbps Ethernet systems by the mid 90s. Copper cable will not allow the speeds attainable with fiber optic cable. However, the standard for fiber stipulates LAN speeds of only 100 Mbps, far below the fiber optic cable's actual capacity.

Twisted pair is grouped into certain classifications based on quality and transmission characteristics. The classifications are called "types" by IBM. UTP by itself is often grouped by "grades".

Unshielded Twisted Pair Grades

Grade 1 Suitable for voice transmission and data transfer up to 1 Mbps

Grade 2 Capable of carrying data at 4 Mbps

Grade 3 Carries data at up to 10 Mbps

Grade 4 Rated at 20 Mbps

Grade 5 Supports speeds at up to 100 Mbps

Twisted Pair Cable Types (IBM Standards)

Type 1 STP, two pair, 22 gauge, solid conductors, braided-shield

Type 2 Type 1 cable with additional four pairs of UTP

Type 3 UTP, 22 or 24 gauge, 2 twists per foot, four pairs

Type 5 Fiber optic cable used to link MAUs

Type 6 Two pair, stranded (not solid) 26 gauge, patch cables

Type 8 Two pair, 26 gauge, untwisted but shielded cable

Coaxial cable or just "coax" enjoys a huge installed base among LAN sites in the US. It has fit the bill perfectly for applications requiring stable transmission characteristics over fairly long distances. It has been used in ARCnet systems, Ethernet systems and is sometimes used to connect one hub device to another in other systems. This is due to coax's superior distance allowances.

Construction-wise coax is a little more complex then TP. It is typically composed of a copper conductor that serves as the "core" of the cable. This conductor is covered by a piece of insulating plastic, which is covered by a wire mesh serving as both a shield and second conductor. This second conductor is then coated by PVC or other coating. The conductor within a conductor sharing a single axis is how the name of the cable is derived.

Figure 2-13: Coaxial cable's use of a second conductor doubling as shielding helps reduce effects of outside interference.

Coaxial cable's construction and components make it superior to twisted pair for carrying data. It can carry data farther and faster than TP can. These characteristics improve as the size of the coax increases. There are several different types of coax used in the network world. Each has its own RG specification that governs size and impedance, the measure of a cable's resistance to an alternating current. One must be cautious in acquiring coax to make sure the right kind has been obtained. Different cable can differ widely in many important areas.

Common Coaxial Cable Types Used In Networking

Type Common Usage Impedance

RG-8 Thick Ethernet 50 ohms

RG-11 Broadband LANs 75 ohms

RG-58 Thin Ethernet 50 ohms

RG-59 Television 75 ohms

RG-62 ARCnet 93 ohms

Twisted pair has one chief advantage, however, and it's an important one. TP is less expensive than coax. In addition, as mentioned in our earlier section, TP is often already available on-site due to phone installation. TP is also extremely flexible and easy to work with, though it may not be as sturdy as coax. Because of these factors, the current marketplace has migrated away from coax and it is no longer the "chic" cable to buy. Plus, most development research is based on improving performance on twisted pair systems. Coax still has specific purposes, which means it won't go away, but its role as primary choice for cabling is no longer accepted in the marketplace.

Great caution should be used when selecting connectors for coax. There is standardly available about 4 different kinds of connectors. The first is the factory pre-molded connections. These tend to be quite sturdy and reliable. The second type is soldered connections. These too appear to be very reliable and durable. The third type is the crimped- on connections. Crimped connections are as good as the crimp tool and the crimper. If a proper tool is purchased (between $75 and $140, US), then getting the crimp right every time should be no problem. There is never any need to apply excessive force to the tool. The worst-case scenario is crushing the cable during a crimp resulting in a short. The fourth kind of connector is called screw-on connector. Screw-on connectors are notorious for being extremely flimsy and unreliable. You do yourself well to avoid them.

All told, coax is an excellent medium for LANs, just expensive in comparison to UTP. Its widespread use will ensure that its existence is supported for quite some time.

Carrying data at dizzying speeds, fiber has come into its own as the premier bounded media for high speed LAN use. Because of fiber's formidable expense, however, you're not likely to see it at the local workstation any time real soon. Instead, fiber is used to link vital components (like file servers) in a LAN or multi-LAN environment together. Consequently we often hear terms like "fiber backbone" thrown around.

Fiber optic is unsophisticated in its structure, but expensive in its manufacture. The crucial element for fiber is glass that makes up the core of the cabling. The glass fibers may be only a few microns thick or bundled to produce something more sizable. It is worth noting that there are two kinds of fiber optic cable commercially available - single mode and multimode. Single mode is used in the telecommunications industry by companies like AT&T or US Sprint to carry huge volumes of voice data. Multimode is what we use in the LAN world.

The glass core of a fiber optic cable is surrounded by and bound to a glass tube called "cladding". Cladding adds strength to the cable while disallowing any stray light wave from leaving the central core. This cladding is then surrounded by a plastic or PVC outer jacket with provides additional strength and protection for the innards. Some fiber optic cables incorporate Kevlar fibers for added strength and durability. Kevlar is the stuff of which bullet-proof vests are made, so it's tough.

Figure 2-14: Fiber optic cable provides tremendous bandwidth for data transmissions. Its construction makes it a very durable medium.

Fiber optic is lightweight and is utilized often with LEDs (Light-Emitting Diodes) and ILDs (Injection Laser Diodes). Since it contains no metal, it is not susceptible to problems that copper wiring encounters like RFI and EMI. Plus, fiber optic is extremely difficult to tap, so security is not a real issue.

The biggest hindrance to fiber is the cost. Special tools and skills are needed to work with fiber. These tools are expensive and hired skills are expensive too. The cable itself is pricey, but demand will ease that burden as more people invest in this medium. Attempts have been made to ease the cost of fiber. One solution was to create synthetic cables from plastic as opposed to glass. While this cable worked, it didn't possess near the capabilities of glass fiber optic, so its acceptance has been somewhat limited. The plastic fiber cables are constructed like glass fiber only with a plastic core and cladding.

The bandwidth or capacity of fiber is enormous in comparison with copper cabling. Multimode fiber can carry data in excess of 5 gigabits per second (that's million megabits). Single mode fiber used in telecommunications has a theoretical top speed in excess of 25,000 Gbps. That much data is the equivalent of all the catalogued knowledge of man transmitted through a single small glass tube in less than 20 seconds. That's impressive.

The standard governing implementation of fiber optic in the marketplace is called the Fiber Distributed Data Interface standard or FDDI. FDDI specifies the speed of the LAN, the construction of the cable, and distance of transmission guidelines. FDDI behaves very much like token ring, only much faster. An added feature for FDDI is a backup ring in case the main ring fails. This fault tolerance along with the fault tolerance already incorporated in token ring technology makes FDDI LANs pretty resilient. One minor drawback for fiber optic LANs is that they can be difficult to layout.

Now that we've examined the major bounded media, let's take a quick look at how they compare.

The dream of being able to communicate data in networks without having deal with the constraints of physical cabling is very much realized today. Wide area networks obviously make use of wireless technology to transmit data around our globe. The acceptance of wireless networks on the local level has been significantly hindered, however, for a number of reasons.

Perhaps the biggest drawback to the two major local wireless technologies - radio and infrared - has been their speed. Neither could come close to matching the 10 or 16 Mbps provided by conventional bounded media LANs. In fact, until recently, these technologies were struggling within their confines to reach out of the Kbps range. Today, however, wireless LANs are climbing out of the doldrums with comparable speeds to token ring systems. The perception that they are slow and limited is still fairly widespread, however, which will limit wireless' acceptance on the desktop.

Additionally, the size of the installed base of physical wiring plays a part in unbounded local media acceptance. The United States, for instance, has a very large installed base of physical cabling. It's readily available and fast. Other countries like Japan, surprisingly enough, do not have such a large installed base. Consequently, their marketplaces are more open to the idea of wireless LANs and emerging higher speed technologies may find better acceptance there.

Another major hurdle for wireless LANs will be the standardization process. This is necessary if there is ever any hope for interoperability in the marketplace between products from different vendors. The IEEE has created a committee that will oversee this standardization. The standard will be called the 802.11 standard.

Radio offers superior characteristics as a wireless media but suffers from a major hindering force known as the government. The government doesn't mean to hinder radio LANs, but the Federal Communications Commission must bridle radio for LAN use in order to responsibly manage our public airwaves, and that is, after all, what we pay them to do. Fortunately, radio LAN product manufacturers have isolated frequencies that are not licensed by the government and made use of these allowing them to scoot under the regulatory fence.

Radio transmitters are omnidirectional and can easily penetrate walls, floors, ceiling and the like. Electrically speaking, the waves that are classified as radio waves have certain frequencies that are grouped together for certain uses. Some are available for data transmission, but the bandwidth necessary to perform high speed data transfers is not found at any given slot on the radio spectrum. Many vendors are now employing spread-spectrum technology where the available slots in the radio spectrum are all used together. Using this technology, speeds at up to 2 Mbps have been achieved.

Figure 2-15: Radio-based LANs use portable transmitters and receivers at each LAN device.

Radio-based LANs do have to contend with the interference that occurs daily in the workplace. That interference can come from a number of different electrical sources and can be quite impacting on LAN performance. For radio systems using only a small portion of the radio spectrum (narrowband systems), this could mean that problem might be insurmountable. The vendors of spread-spectrum products claim that their products can isolate interference problems and avoid using those frequencies.

Though radio offers portability to any node within range, its unbounded nature makes it somewhat less secure. A "non-friendly" could, in theory, listen in to your radio broadcasts. The eavesdropper would have to, of course, know what frequency or frequencies you were using. Once that hurdle was overcome, your LAN would be laid bare.

Radio, though limited by its speed, may be the wireless transmission method of choice for many desktops because of its low cost and capabilities. However, the delay of regulation has cost radio a few months before standardization. This has given infrared vendors at least a little time to create competing products.

Infrared technology uses the invisible portion of the light spectrum with wavelengths just a little less than those of red light. These frequencies are very high offering nice data transfer rates. Modern infrared LANs can achieve throughput at 16 Mbps with potential for better. We are used to seeing infrared technology utilized for our television or VCR remotes.

Figure 2-16: Infrared transmissions offer potential for high speed data transfer but are limited by inability to penetrate walls and floors.

Infrared technology involves the use of an infrared transmitter like an LED or ILD along with a receiver, typically a photodiode. These components operate in a line-of-sight fashion. That is, nothing can obstruct the pathway between them. Fortunately these signals can be bounced off walls and ceilings providing transmission around obstacles. Line-of-sight means, however, that these signals cannot be broadcast through walls, severely limiting infrared LANs.

Modern infrared systems use a repeater device simply to retransmit a signal from one room into another. This device is generally mounted on the ceiling or high in a corner to alleviate as many obstacles as possible. These systems also use a process called "diffusion" to send the signal in a wide path across a room thus reducing the chance of signals not getting past a single obstacle.

The good news about infrared technology is that it may not be very costly to implement. Since infrared items have been around a while, significant resources exist to mass produce infrared products. Advances in the technology will probably lead to faster products without as many limitations. Infrared transmissions now are limited to a relatively short distance, and used outdoors, are extremely susceptible to atmospheric conditions.

1. Know the various topologies and their characteristics including layout and troubleshooting considerations.

 

2. Know the difference between a physical topology and a logical topology. Give an example of each.

 

3. Know which topologies are commonly used for popular LAN types such as Ethernet and token ring.

 

4. Be familiar with the major components of most modern LANs including servers, workstations, print servers, etc.

 

5. Know the major communication schemes and be able to compare them according to traffic conditions.

 

6. Know what CSMA/CD is.

 

7. Be familiar with the major types of cabling available and the construction of each.

 

8. Compare the cabling mediums as far as cost, performance and susceptibility to interference.

 

9. Know the two major types of wireless media and their characteristics.

 

Table of Contents   Chapter 1   Chapter 3   Netguru.net   Glossary

Copyright© 1993 & 1997 by Atrium Technical Inc. All rights reserved. Illegal to copy without written permission of ATI.