'; zhtm += ''; zhtm += ''; zhtm += '

' + pPage + ''; zhtm += ''; window.popUpWin.document.write(zhtm); window.popUpWin.document.close(); // Johnny Jackson 4/28/98 } //--> Managing Multivendor Networks -- Ch 7 -- LANs and WANs

/www.cisco.com/

LANs and WANs

The Many Faces of Networking

efore the great LAN explosion, networking, for the most part, addressed the connection of distributed devices to a central location. Although some pioneering companies, such as Digital Equipment, offered LAN technology in these early days, the bulk of the market was accustomed to a centralized computing environment.

In this centralized approach, the primary concern was to find the most practical and economical way to connect terminals, printers, and other data collection/reception devices to the primary location. When connectivity was required between systems, the link was approached typically as a special-case, point-to-point operation, rather than part of a peer-oriented, distributed processing network. However, as requests mounted to link computer systems over wide areas, multiple, point-to-point operations became very cost ineffective, and the door opened to such alternative wide-area connections as X.25 and ISDN. Wide-area technologies have continued to evolve, and now include Frame Relay, Asynchronous Transfer Mode (ATM), and Switched Multimegabit Data Service (SMDS).

All things considered, this system-to-system connectivity hardly concerned the end user--after all, this was the job of the communications analyst. But when the LAN wave finally reached the PC on the end user's desk, that user suddenly encountered and became concerned about connectivity issues. At first it was just local (LAN) connectivity and terminal emulation. Then, as networks grew and costs increased, products such as gateways, bridges and routers snaked their way into the LAN. Today, the end user has an unprecedented amount of power at his or her disposal. Consolidated, enterprise-wide data is no longer in the hands of a few technical elite; off-the-shelf desktop software now gives the end user the ability to access data anywhere in the enterprise--whether it is on the PC, server, minicomputer, or mainframe.

This progression of connectivity changed the role of the LAN. Whereas the LAN began as a local computing environment (usually an island unto itself) it grew into an area of computing, normally linked to other computing areas. The fact that one computing area might be a LAN, another a mainframe, and yet another a combined midrange computer and PC LAN has become almost irrelevant.

From this high-level perspective, the world of distributed networks can be broken down into two large categories: local-area links and wide-area links. A local area network (LAN) typically is limited to one geographic area and allows individual workstations to access data or applications on a server. In smaller LANs, a peer-to-peer arrangement can be deployed to allow each station to function as both server and client. A wide area network (WAN), on the other hand, typically covers a large geographic area, and often links together multiple LANs. Within each category, however, are a wide variety of implementations and strategies.

LANs often play a pivotal role in modern networks. This chapter will address the following LAN issues:

WANs can be used in both centralized and distributed processing environments to tie all of the necessary devices together. This chapter will discuss the following WAN issues:

Local Area Networks

LANs became significant in the world of networking in the late 1980s, following on the heels of the PC to become the preferred method for connecting multiple PCs in a self-contained area.

Unfortunately, the networking software and operating systems used with the PC LANs were quite different from the networking software used on midrange and mainframe computers and office automation equipment (for example, dedicated word processing machines and intelligent copiers). This, of course, set up the inevitable conflict between PCs, office automation equipment, and the larger midrange and mainframe computers. Although many computer-savvy corporations saw the conflict coming and took steps to address it head on, other companies first became aware of the conflict when the requests to run cable hit the maintenance department.

After all, from a simple and fundamental perspective, the laying of the cable represents a major commitment. Installing the cable requires the unpleasant work of snaking cable through ceilings and down walls. It requires that the cable be arranged in such a way that it is manageable and easy to expand (from a networking perspective). And even worse, the placing of cable is often regulated by local ordinances that require special casings or materials (Teflon enclosures, for example) if the cable runs near pipes, electrical work or people. In short, putting the cable in is almost as much fun as simultaneously remodeling the kitchen and bathroom of your house.

LAN Topologies

Whether the purpose of the LAN is to interconnect PCs, minicomputers, or both is almost irrelevant--the first issue is often choosing the topology of the LAN. This choice dictates the cable, cabling methodology and the networking software that can operate on the LAN. The three basic topologies are the ring, star, and bus (see Figure 7.1).

FIG. 7.1 LAN Topologies

Star and ring network topologies are sometimes combined into one network to provide a higher degree of fault tolerance. Because a star network is susceptible to a failure in the hub, and a ring network is sensitive to a break in the ring, combining both forms into one offers an alternate route in case one topology fails.

LAN Protocols

While the LAN topology defines the cabling methodology and the way that information flows through the network, the LAN discipline determines how the computers interact with each other on the LAN. The two most used protocols are token passing and collision sensing.

In a token passing network, a special token is passed from computer to computer. Possession of the token enables a computer to transmit on the network. When the original transmission returns to the computer that sent it, that transmission is regarded as complete (whether or not it was actually received) and a new token is generated to flow to the next station (based on the LAN topology). When a message is successfully received, the receiving station confirms receipt by changing a flag in the original transmission. Therefore, by examining the original message when it returns, the sending station can determine what happened at another end of the ring. Token passing dates back to 1969 and is one of the earliest multiple-unit, peer-to-peer control procedures. Token ring networks, although not as widely used as a CSMA/CD network, offer robust performance because they furnish only a single channel; thereby avoiding any possibility of collision.

The proper name for collision sensing is Carrier Sense Multiple Access with Collision Detection (CSMA/CD). With this discipline, each computer listens to the LAN to sense if another computer is transmitting. If someone else is active on the LAN, the computer wishing to transmit waits for a preset amount of time before trying again. When the computer perceives that the LAN is inactive, it transmits. In the event that two computers transmit at the same time (and their data collides and is hopelessly corrupted), both sides wait for different lengths of time before attempting to retransmit. CSMA/CD dates back to the mid-1970s (when Ethernet was in its infancy) and has grown to be the most common discipline for PC LANs.

The biggest difference between the two disciplines is that token passing is termed a deterministic discipline while collision sensing is not. A token passing network is deterministic because each computer is given the opportunity to transmit, but only at preset time intervals, and only if it is in possession of the token. On a collision sensing network, however, each computer must, in effect, compete for the opportunity to transmit.

A third type of discipline, time division, is sometimes used in laboratory environments for specialized controllers, technical equipment or wireless communications networks. With this discipline, each unit is given specific amounts of time at specific intervals to exchange data. Using time division in a conventional data processing LAN, however, is extremely unusual, not to mention impractical.

LAN Implementations

Both the discipline and topology define a LAN implementation. Thus, a LAN might be a token passing ring, a token passing bus, or a collision sensing bus.

The standards for LAN implementations can be roughly broken into two groups: those that pre-date the work performed by the IEEE in this area, and those that were developed by the IEEE. Of the LAN implementations that pre-date IEEE's involvement, Ethernet and token ring implementations have stood the test of time and remain popular.

Ethernet was originally developed by Xerox Corporation in the 1970s as a 3 Mbps bus LAN using the CSMA/CD discipline. Following the initial release of Ethernet, both Digital Equipment and Intel joined the development effort and the three companies released the specification for Ethernet version 1.0 in 1980. The most notable improvement in Version 1.0 was the increase in the LAN speed from 3 Mbps to 10 Mbps. The Ethernet specification was then revised again several years later as Ethernet II to provide a higher degree of compatibility with the IEEE 802.3 standard. The 802.3 standard has since grown to include a newer specification, known as Fast Ethernet, which boosts the speed tenfold to 100 Mbps.

In addition, the IEEE 902.9a isochronous Ethernet standard provides a way for two networks to run over 10Base-T wiring. IsoEthernet permits the integration of LAN and WAN services, and can extend a company's existing investment in standard Ethernet. IsoEthernet can deliver voice and video as well. In the past, multimedia over Ethernet has been limited because of Ethernet's connectionless nature. Traditional Ethernet generates bursty traffic, which is excellent for sending data, not suitable for time-sensitive information such as video. This type of time-sensitive traffic is highly dependent on all packets arriving in the correct order. IsoEthernet is capable of multiplexing 56 Kbps/64 Kbps ISDN B channels and running both packet and wideband circuit-switched multimedia services over Category 3 UTP cable. Its encoding scheme also increases the available bandwidth from 10 Mbps to 16 Mbps. The extra 6 Mbps of bandwidth is used to create a multimedia pipe. IsoEthernet can be integrated into an existing 10Base-T Ethernet with the addition of an isoEthernet hub, which permits WAN and LAN services to be synchronized. Workstations must be equipped with isoEthernet adapter cards, which are connected to the hub. An attachment unit interface (AUI) then connects the isoEthernet and Ethernet hubs.

Token ring networks have been implemented on a variety of media at a variety of speeds. Therefore, unlike Ethernet, token ring technology was not successfully introduced into a generalized data processing network. IBM implemented token ring in its early PC LANs, as did Apollo for its engineering workstations. But somehow, token ring did not catch on as Ethernet did. There were several reasons for this, including the fact that token ring is more expensive to deploy than Ethernet, requires more planning, and is more difficult to install. More recently, however, token ring networks have enjoyed a rebirth in popularity for several reasons.

The IEEE organization adopted token ring as a sanctioned network in its IEEE 802.5 spec-ifications.

The market for token ring switches is enjoying tremendous growth as corporate networks continue to grow at an unprecedented pace. These switches provide users on overcrowded LANs with their own personal 4 Mbps to 16 Mbps piece of bandwidth. The switch can also be used to divide a large ring into smaller segments.

Until recently, equipment for switched token ring networks was largely unavailable. However, token ring networks can suffer from the same geographic limitations as Ethernet, and vendors are now stepping in to provide the switching equipment users require to expand their token ring networks. Traditionally, two-port bridges are used in token ring networks, which impose a significant limitation on its expandability. Token ring switches can connect the separate rings to each other and to servers, without the performance limitations of the past. Some products include both token ring and Ethernet switching facilities in the same box. Most switches also accommodate high-speed networking, such as ATM or FDDI; many also support RMON management.

Both token ring and Ethernet networks have bandwidth limitations. Many corporate networks are beginning to reach those limitations, as they bring in more and larger applications and experience a greater demand for data. Switching technology can help overcome these limitations by extending an overcrowded network. Whether the switch is used to divide the ring into smaller segments or to give each user a personal slice of bandwidth, switches can greatly enhance network performance, thereby extending the useful life of the existing network.

IBM has made a significant commitment to supporting token ring as the preferred SNA LAN and has, in fact, provided connections for its broad range of computers and communications controllers to token ring.

The architecture of the Fiber Distributed Data Interface (FDDI) is modeled after token ring. FDDI is a high-speed WAN technology that runs at 100 Mbps. Therefore, if you love FDDI, you must also at least have a passing respect for token ring.

Token ring technology was patented by a European engineer who forced those who adopted it to pay a royalty. However, this patent has been successfully challenged, so the economics of token ring networks has taken a turn for the better.

LAN Implementation Standards

Although both Ethernet and token ring networks function well, they were not recognized as official standards because they were developed in the private, commercial sector. To address this need for standardization, the IEEE studied these and other implementations and developed a series of standards to properly define a series of LAN specifications.

In developing its standards, the IEEE had to walk the line between the OSI Reference Model and the existing, well-known and widely accepted LAN implementations. In terms of the OSI Reference Model, for example, IEEE carved the Data Link Layer (layer 2) into two parts. The upper half of the layer that interfaces with the Network Layer (layer 3) was termed the Logical Link Control (LLC). The LLC provides a common, low level point of access, independent from the actual physical media.

The lower half of the layer that interfaces with the Physical Layer (layer 1) was termed the Medium Access Control (MAC). The MAC addresses the specifics of the physical network interface; therefore, separate MAC standards are defined for CSMA/CD, token passing bus and token passing ring. However, note that a single LLC specification addresses all three MACs.

A message passed from the Network Layer is processed by the LLC protocol, and an LLC header is added to the data (see Figure 7.2). This new data structure is then passed on to the MAC where another header and a trailer are added before the data enters the physical network. The resulting structure that includes the MAC header, the LLC header, the data and the MAC trailer is termed a frame.

IEEE 802.2

In IEEE terms, the 802.2 specification defines the LLC (see Figure 7.3). The 802.2 header consists of the following:

Below the 802.2 LLC are the MACs for the various physical LAN implementations. These standards are known as 802.3 for CSMA/CD, 802.4 for token passing bus, and 802.5 for token passing ring.

FIG. 7.2 IEEE LLC and MAC Layers

IEEE 802.3

The IEEE 802.3 standard specifies a CSMA /CD bus network that supports 10-Mbps transmission over baseband, broadband, and twisted pair cable. This networking standard closely resembles Ethernet. Both HP and IBM (and others) support the IEEE 802.3 networking standards (HP for their native NS networking product, and IBM for their TCP/IP products).

The 802.3 header (see Figure 7.4) includes the following:

FIG. 7.3 IEEE 802.2 LLC Header

Following the header is the 802.2 header and the actual data. At the end of the data is the 802.3 trailer, which includes:

IEEE 802.4

The IEEE 802.4 specification defines a token passing bus that can operate at speeds of 1, 5, or 10 Mbps. The 802.4 standard is, in many ways, a marriage of Ethernet and token ring technologies. The physical topology for 802.4 is a bus, much like in Ethernet, but the MAC-level discipline is a token-passing logical ring (as opposed to a token-passing physical ring). Although the 802.4 specification does not have as many active supporters as the 802.3 and 802.5 standards, its popularity is rapidly growing. The format for 802.4 transmissions (see Figure 7.5) is as follows:

FIG. 7.4 IEEE 802.3 CSMA/CD Frame

Following this header is the 802.2 header and the actual data. At the end of the data is the 802.4 trailer, which includes the following:

IEEE 802.5

The IEEE 802.5 standard specifies a token passing ring operating over shielded twisted pair cables at speeds of 1, 4, or 16 Mbps. This standard is supported by IBM in its Token Ring implementation. The 802.5 construction (see Figure 7.6) is defined as follows:

FIG. 7.5 IEEE 802.4 Token Bus Frame

FIG. 7.6 IEEE 802.5 Token-Ring Frame

Following this header is the 802.2 header and the actual data. At the end of the data is the 802.5 trailer that includes the following:

802.3 Versus Ethernet

The implementations of Ethernet and IEEE 802.3 are so compatible that computer systems using each can coexist on the same network. The most significant difference between the two is the way information is formatted into frames. Although both specifications define the destination and origin of the information, the 802.3 frame includes significantly more detail.

The Ethernet frame begins in the same fashion as the 802.3 frame with a preamble, start delimiter, and then the destination and source addresses (see Figure 7.7 ). The similarity stops here, because in Ethernet these addresses are followed by a type field, which identifies which Ethernet service the frame applies to. However, because the headers are so similar, these frames can coexist on the same LAN without interfering with one another ( providing that the 802.3 frame uses 48-bit addresses as does Ethernet).

FIG. 7.7 Comparison of Ethernet and 802.2/802.3 Frame Formats

NOTE: Ethernet Version 1.0 does not have the same level of compatibility with IEEE 802.3 as Ethernet II. Specifically, the primary difference is that Ethernet II and IEEE 802.3 both include a "heartbeat" function performed by the transceivers (units that attach computer and computer equipment to the physical LAN) to signal their ongoing operation (the absence of a heartbeat signals a failed or failing transceiver).

Wide Area Networks

At a basic level, a WAN can be created by tying a series of simple, point-to-point links together. On the other end of the spectrum, a WAN might comprise many different systems and LANs, all interconnected using a variety of techniques, including standard telephone lines, packet-switching networks and ISDN links. Between the two extremes are networks that are superficially simple but technically complex, and those that are superficially complex but technically simple.

Unlike LANs, which all accomplish the same purpose, WANs offer a unique variety of technology and approaches. This discussion will focus on some of the better known approaches--standard phone links, ISDN networking, and X.25 packet-switching--as well as emerging methods such as ATM, Frame Relay, and SMDS.

Like LANs, WAN links are simply a way of transferring information from point A to point B. Running on top of both types of links are networking protocols and services that bring additional functions to the network. For example, IBM's SNA, Digital's DECnet, TCP/IP, and many other networking protocols all include services that operate over the physical links. Some protocols are specific to the LAN environment (such as Digital's LAT or Novell's IPX), while other protocols are better suited for wide area links (like IBM's SDLC or HP's implementation of HDLC).

The point is, in all cases, no network (wide or local) provides any value without upper layers of protocols, services and applications.

Point-to-Point Links

In most cases, long-distance point-to-point links are routed through a telephone carrier. From a practical point of view, the long-distance telephone carriers have already done the work of establishing a wide area of physical links, so it makes sense in some circumstances to use these existing connections.

Before the advent of high-speed digital lines, this world of long distance teleprocessing was composed of dial lines and leased lines. Dial-up POTS (Plain Old Telephone Service) lines are one of the few aspects of data communications that has not changed much over the years, although higher-speed modems have enabled data to be sent over them much faster. As the name implies, a dial line uses standard voice-grade lines to create a temporary connection between two computing devices. POTS lines can operate at speeds of up to 36,600 bps if a noise-free connection can be made.

A leased line is a permanent circuit installed between point A and point B. Because they are permanent, leased lines can be conditioned to provide less noise and therefore support high-speed operation (such as 28,800 bps) on a more reliable basis. Leased lines have been greatly affected by the advent of digital phone circuits.

Before the advent of digital lines, point-to-point links used the same basic approach to carry data as they did to carry voice. Although, as noted, leased lines could be purchased with various levels of conditioning, they still used the same analog approach for transmitting. Modems were developed to bridge the difference between the analog nature of the phone system and the digital nature of computers. Modems that translate between the digital and analog formats are described in greater detail at the end of this chapter.

NOTE: The analog /digital translation process, MOdulating and DEModulating, forms the etymology of the word "modem."

As technology and phone systems matured, the nature of the phone network became much more sophisticated. Satellites were deployed to provide greater coverage without costly physical connections, and digital circuits were added into many phone systems to offer high transmission speeds with lower noise (and therefore fewer errors).

In particular, Digital Data Service (DDS) brought increased performance to leased lines. When compared to digital networking, the analog phone system is slow and error-prone. Furthermore, because the existing phone system was developed to address voice transmissions, the way it handles data communications is less than ideal. Digital service brought increased reliability and performance to leased line networks.

Higher rates are provided through the use of T1 links. T1 links are multiple, high-speed links packaged into a single unit. Specifically, a T1 line has an aggregate throughput of 1.544 Mbps but is, in reality, composed of 24 64-Kbps digital lines.

A T1 user can dedicate these 24 lines to separate functions--for example, some might carry voice, some video and some data. Or a T1 user can use multiplexing equipment to run data across all (or a subset) of the separate lines concurrently, to effectively achieve the full throughput. If a company does not need the full T1 bandwidth, it might also choose fractional T1 service. In this case, only some of the T1 lines are connected to the customer's premises. The availability of fractional T1 lines is dependent on the local phone company's ability to find enough fractional users to use up an entire T1 link.

Through bridges and routers, geographically distant LANs can be interconnected over a T1 link. However, since T1 is strictly a North American standard, it cannot be used to establish an intercontinental WAN.

NOTE: Because of the wide and diverse geography of the U.S., digital and T1 services are not available in all areas of the country. 

ISDN

ISDN is the planned replacement for the analog circuits used to provide voice and data communications services worldwide. Development on ISDN was begun in the 1970s by AT&T and formalized in the early 1980s under the mantle of the Consultative Committee for International Telegraph and Telephony (CCITT). Under the direction of CCITT, ISDN became worldwide in scope, offering for the first time a fixed set of interfaces and interface devices that were applicable globally. Thus, the computer interface used in Germany for ISDN attachment would be the same interface used in the U.S.

The customer's interface to ISDN is through a service node (see Figure 7.8). The purpose of the service node is to provide an interface from ISDN to the customer phone system or PBX, a data communications device that interfaces to the local computer equipment (analogous to a modem or CSU/DSU), or a hybrid device that performs both functions. This service node interface enables the customer to access ( given proper security and compatible equipment) any other system also connected to ISDN.

FIG. 7.8 ISDN Service Node Concept Interface

For data communications, this approach is more flexible than the traditional point-to-point leased line or dial-up connections previously discussed. Voice service, on the other hand, will not be dramatically improved because it also operates at the equivalent of 64 Kbps over the analog system. The digital nature of the network should, however, remove some of the static often heard over phones. (Also remember that while static is annoying to humans, it is devastating to data, and that's why data is not transmitted across analog lines at these extremely fast speeds.)

From a point of entry perspective, ISDN offers two types of user interfaces:

Because the costs for a PRI far exceed the cost for a BRI, most business needs are addressed by one or more BRIs. Still, given the higher speeds offered by the PRI (a total of 1.544 Mbps for the PRI versus a total of 144 Kbps for the BRI), the PRI is a viable contender to extending LANs using bridges and routers.

Also note that the PRI closely resembles a T1 link. The primary difference lies in their use and network architecture. T1 is most often used to facilitate high-speed point-to-point links, whereas ISDN is intended to interface a large number of systems on a global basis. ISDN also differs from the T1-style link in that it features a management circuit separate from the data circuits. This additional circuit is present in both the Basic Rate and Primary Rate Interfaces and delivers some benefits that are important to ISDN and its marketability. This circuit is separate from the data/voice channels (see Figure 7.9). In fact, it is termed a D Channel, as opposed to the B channels that carry the data and voice traffic.

FIG. 7.9 ISDN B and D Channels

This type D circuit can be used for a number of functions:

Beyond establishing a digital international data network, ISDN is a key piece of the standards pie for several reasons:

Laptop PC users now also have the ability to access a network with ISDN services through new ISDN Basic Rate Interface PCMCIA cards.

The various Regional Bell Operating Companies (RBOCs) are now providing ISDN service to most major cities, and the number of ISDN lines has increased dramatically over the past few years. Additionally, all of the major commercial on-line services offer ISDN access, and many Internet service providers are also offering ISDN Internet access. The advantages are obvious; file downloads are lightning fast, and there is less possibility of interruption due to line noise. PC vendors are accommodating the increasing demand for ISDN by releasing ISDN modems, also known as ISDN terminal adapters.

Although the cost of an ISDN connection is bound to decrease, costs vary tremendously from region to region.

Asymmetric Digital Subscriber Line (ADSL)

Although ISDN has been getting most of the press, a similar technology called asymmetric digital subscriber line (ADSL) promises even more throughput over an ordinary, narrow copper telephone line. ADSL accomplishes this remarkable feat through a series of complex compression and digital signal processing algorithms, and dynamic switching techniques. The ADSL transport technology boosts the capacity of the existing phone line significantly more than ISDN. Duplex ADSL offers a downstream data rate of up to 6 Mbps and an upstream channel running at 640 Kbps. ISDN, on the other hand, ranges from 64 Kbps to 128 Kbps--faster than a standard modem, but still too slow to handle that interactive TV and other services they keep telling us we'll all have one of these days.

Most of the regional Bell operating companies are testing ADSL and making plans to offer it to customers seeking high-bandwidth Internet access. Also, because it offers two-way communications, some entertainment companies are considering it a realistic possibility for interactive cable television.

ADSL gives the RBOCs an alternative to costly optical cables because it can transform their existing copper-wire network into a high-performance system. Here's how it works: The regular phone wire is configured for ADSL, and then connects to an ADSL modem on one end and an ADSL circuit switch on the other. The connection then creates:

Cable companies, set-top box makers, and TV couch potatoes drool over the possibilities. Look for this technology to be making big news in the near future.

X.25

The CCITT developed the X.25 standard to define a reliable, relatively low cost means of routing data through a shared network. An extremely important aspect of X.25 is that the information being transmitted has been converted into packets.

Packets can be thought of as small fragments of information. Specifically, a block of information is broken into smaller parts ( packets) before being transmitted on the physical network. The packet methodology provides faster and more reliable error detection and correction; it also prevents a system with a huge volume of information to ship from tying up the network.

In addition to the raw information, each packet also contains information specifying its origin, its destination, and a number indicating the "piece" of the information to which it corresponds. This enables each packet to be treated as an independent entity, so that packets from many different systems can be intermixed on the network without concern about the order in which they are transmitted or even the order in which they arrive. Each packet might take the best possible route available at the time it is ready for transport.

The application end points of the information (that is, the terminal user and the application program) rarely see the information in its packetized form. As part of its interface with the network, the computer system converts the information into packets, and then subsequently reassembles the packets into the original information (see Figure 7.10).

FIG. 7.10 Conceptual Packetizing

This packet approach to transmitting data is extremely pervasive in the networking world. In addition to being used by X.25, this approach is also used by most LANs and many other data communications protocols (although they are usually referred to as frames, as discussed in the LAN section of this chapter). Specific to X.25 networks, however, is the concept of a packet switching network (PSN).

A PSN is a WAN through which packets are sent. The precise route that packets take from point A to point B is not fixed and is immaterial to the equipment at point A or point B, which checks only to see whether the packets arrive intact (again, order is not a major concern).

Because they don't have prescribed data routes, PSNs are often shown as clouds in many networking diagrams (see Figure 7.11). When depicted in this manner, information goes into the cloud at some point and comes out at another. What goes on within the cloud is not the concern of mere humans.

FIG. 7.11 Typical X.25 Representation

The inside of the cloud, however, is composed of packet-switching nodes (also called PSNs, just to make life confusing). The switching nodes can take routes to other switching nodes, and thus can route or reroute data as necessary. For example, if a switching node has a packet to forward and the best possible switching node to receive it is busy, the node holding the packet will reroute it to another node for subsequent rerouting (see Figure 7.12).

FIG. 7.12 Inside of the X.25 "Cloud"

Packet-switching networks often are associated with public data networks (PDNs), but this relationship is certainly casual. A PDN is normally a telephone system (or telephone company in the U.S.) that offers data services to the public. It does not have to use packet-switching to move information from point to point. If a PDN does offer the services of a packet-switching network, it might be referred to as a packet-switching data network (PSDN) or even a packet-switching public data network (PSPDN). Clearly, the abbreviations are almost endless.

Furthermore, implementation of packet-switching networks is not limited to telephone companies. In fact, PSNs can be constructed of telephone links, fiber optic links, microwave links, satellite links, and other forms of communications. Many large corporations have used these diverse communication techniques to construct their own private PSN. Because, in the final analysis, a packet switching network is a cost-effective WAN, organizations with widely dispersed equipment find this approach most effective in terms of both cost and function.

The traditional packet-switching cloud is shown in Figure 7.13.

FIG. 7.13 X.25 Interfaces

Moving outside of this cloud, the interfaces between the computer equipment and the cloud generally fall into one of two types of devices:

For terminal traffic over packet-switching networks, two additional standards come into play. First, the CCITT X.28 standard defines the interface between an asynchronous terminal and a PAD. Second, the CCITT X.29 standard defines the control procedures for information exchanges between a PAD and another PAD (or an integrated controller). Just as X.25 has become synonymous with packet-switching networks, X.29 has become synonymous with interfacing terminals over packet-switching networks.

LAN Switches

Switches are used to extend overcrowded networks by providing each end user with his own piece of 4 Mbps or 16 Mbps bandwidth. In many cases, this might be more than each end user needs. In this event, the token ring switch can be used to break one big token ring into multiple, smaller rings. This approach will also significantly increase performance.

Switch Technology--Token Ring and FDDI
Many vendors are bringing token ring LAN switches to the market. A number of alliances illustrate the strength of this market, such as a recent noteworthy alliance between Bay Networks (Santa Clara, California) and IBM. Other network vendors, such as Cisco Systems and Cabletron Systems, have made similar deals with third parties. As more vendors go into this market and volume increases, token ring switch products are expected to come down in price and enjoy higher demand.

FDDI switching is another promising technology for extending network life and bandwidth. Digital's GigaSwitch is the leading FDDI switching product, although several other vendors are preparing to release FDDI switches as well.

Token ring networks, like Ethernet networks, have bandwidth limitations, and many are starting to reach those limitations because of the bigger applications and greater demands for data that companies are experiencing. The lower pricing structures of Ethernet and token ring LAN switching devices might encourage individual business units to make their own purchases. In terms of the overall enterprise, however, this can be disastrous. It is essential for individual departments to consider the overall corporate direction when making such purchases, and to make sure that the technology they are purchasing is compatible with the existing infrastructure and corporate data needs assessments. If not, they might wind up spending much more money because they now have to buy additional equipment to connect with the corporate switches and to address data type and volume transmission requirements decided on by corporate information communication needs.

Tools of the Trade

Needless to say, computers and networks do not connect to each other as easy as phones plug into wall jacks. In networks, the tools of connectivity handle conversion between analog and digital formats, between one type of physical interface and another, or between one transmission media and another. In short, these tools are the nuts and bolts of the erector set called networking.

For LANs, one set of tools is required to make the physical attachment between the interface in the computer (for example, an Ethernet adapter in a VAX or a token ring adapter in an AS/400) and the physical network. The tools include:

When two LANs are joined together, a bridge or router is normally used. When a bridge links two or more LANs, those LANs form a single, logical LAN. In this case, all information routed through one LAN goes over the bridge and through the attached network. Because of this traffic, high speed links are normally required to keep the bridge from slowing the performance of the network. And finally, because bridges are implemented at such a low level, all protocols can operate over a bridge.

A router also connects two or more LANs, but routers are much more selective about the information that they allow to cross over. Specifically, routers are aware (through self-learning or manual configuration) of which computer addresses apply to which LANs. Therefore, rather than pass all information across, routers transmit only information pertinent to the other LAN. Because only selected traffic travels across the link, lower speed links can be used without affecting overall LAN performance. The router can also act as a firewall to prevent unwanted access to the network from outside.

Routers cannot be used in all types of networks, though. Because routers depend on the network to supply an internetwork address (an address that is globally unique), those network protocols that do not support this type of addressing cannot be used with routers. Digital's LAT protocol, for example, has no facilities for internetwork addressing, and therefore will not travel over a router (but it will travel over a bridge). And because routers and bridges have their advantages and disadvantages, the two are often combined into one piece of equipment (in this case some protocols are bridged and others are routed). These devices are often called brouters.

When a computer, bridge, or router must interface to the telephone system (analog, digital or T1), more special devices are needed. They are as follows: