When the data arrives at its destination, the receiving computer performs the same procedure as the transmitting computer, except in reverse. The data is passed up through the layers to the receiving application, with each protocol providing an equivalent service to the protocol in the layer above it. For example, if a protocol at layer three on the transmitting computer is responsible for encrypting data, the same protocol at layer three of the receiving system is responsible for decrypting it. In this way, protocols at the various layers in the transmitting system communicate with their equivalent protocols operating at the same layer in the receiving system, as illustrated in Figure 1.2.

Figure 1.2   Protocols operating at the same layer in the stack on different systems can be said to communicate indirectly by providing complementary services

Local Area Networks and Wide Area Networks

A LAN is a group of computers located in a relatively small area and connected by a common medium. Each of the computers and other communicating devices on the LAN is called a node . A LAN is characterized by three primary attributes: its topology, its medium, and its protocols. The topology is the pattern used to connect the computers together. With a bus topology, a network cable connects each computer to the next one, forming a chain. With a star topology, each of the computers is connected to a central nexus called a hub or switch. A ring topology is essentially a bus network with the two ends joined together. You will learn more about network topologies in Chapter 2, "Network Hardware."

The network medium, as defined earlier, is the actual physical connectionbetween the networked computers. The topology and the medium used on a particular network are specified by the protocol operating at the data-link layer of the OSI model, such as Ethernet or Token Ring. Ethernet, for example, supports several different topologies and media. When you select one combination oftopology and medium for a LAN, such as unshielded twisted pair (UTP) cable in a star topology, you must (in most cases) use the same topology and medium for all of the computers on that LAN. There are some hardware products that enable you to connect computers to the same LAN with different media, but this is only true for closely related technologies. You can't connect a bus Ethernet computer to a star Ethernet computer and have both systems be part of the same LAN.

In the same way, all of the computers on a LAN must share common protocols. You can't connect an Ethernet computer to a Token Ring computer on the same LAN, for example. The same is true for the protocols operating at the other layers of the OSI model. If the systems on the LAN don't have common protocols at every layer of the stack, communication among them is not possible.

In most cases, a LAN is confined to a room, a floor, or perhaps a building. To expand the network beyond these limits, you can connect multiple LANstogether using devices called routers. This forms an internetwork , which isessentially a network of networks. A computer on one LAN can communicate with the systems on another LAN because they are all interconnected . By connecting LANs in this way, you can build an internetwork as large as you need. Many sources use the term network when describing a LAN, but just as many use the same term when referring to an internetwork.

It is important to distinguish between an internetwork, which is any collection of interconnected LANs, and the Internet. The Internet is the ultimateexample of an internetwork, but not every internetwork involves the Internet.

In many cases, an internetwork is composed of LANs in distant locations. To connect remote LANs, you use a different type of network connection: a WAN connection. WAN connections can use telephone lines, radio waves, or any one of many other technologies. WAN connections are usually point-to-point connections, meaning that they connect only two systems. They are unlike LANs, which can connect many systems. An example of a WAN connection would be a company with two offices in distant cities, each with its own LAN and connected by a leased telephone line. This type of WAN is illustrated in Figure 1.3. Each end of the leased line is connected to a router and the routers are connected to individual LANs. Any computer on either of the LANs can communicate with any one of the other computers at the other end of the WAN link or with a computer on its own LAN.

Figure 1.3   WAN connections create internetworks by connecting LANs in distant locations

Broadband and Baseband

In most cases, LANs use a shared network medium. The cable connecting the computers can carry one signal at a time, and all of the systems take turns using it. This type of network is called a baseband network. To make a baseband network practical for many computers to share, the data transmitted by each system is broken into separate units called packets . If you were to tap into the cable of a baseband network and examine the signals as they flow by, you would see a succession of packets generated by various systems and destined for various systems, as shown in Figure 1.4. When your computer transmits an e-mail message, for example, it might be broken into many packets, and the computer transmits each packet separately. If another computer on the network also wants to transmit, it would also send one packet at a time. When all of the packets constituting a particular transmission reach their destination, the receiving computer reassembles them back into your original e-mail. This is the basis for a packet-switching network.

Figure 1.4   In a packet-switching network, the medium carries a seemingly random stream of packets generated by various computers on the network

The alternative to a packet-switching network is a circuit-switching network , in which the two systems wanting to communicate establish a path through the network that connects them (called a circuit ) before they transmit any information. That circuit remains open throughout the life of the exchange and is broken only when the two systems are finished communicating. This is an impractical solution for computers on a baseband network, because two systems connected by a circuit could conceivably monopolize the network medium for long periods of time, preventing other systems from communicating. Circuit switching is more common in environments like the Public Switched Telephone Network (PSTN), in which the connection between your telephone and that of the person you're calling remains open for the entire duration of the call.

To make circuit switching practical, telephone companies use broadband networks. A broadband network is the opposite of a baseband network, in that it carries multiple signals in a single cable at the same time. One broadband network that you probably use every day is the one operated by your local cable television company. A cable TV service runs a single cable into a user 's home, but that one cable carries the signals for dozens of TV channels simultaneously, and often provides Internet access as well. If you have more than one TV in your home, the fact that you can watch a different program on each proves that the one cable carries multiple signals at the same time. Broadband technologies arealmost never used for local area networking, but they are becoming an increasingly popular solution for wide area networking.

Half-Duplex and Full-Duplex Communications

When two computers communicate over a LAN, data typically travels in only one direction at a time, because the baseband network used for most LANs supports only a single signal. This is called half-duplex communication. By contrast, two systems that can communicate in both directions simultaneously are operating in full-duplex mode. These two kinds of communication are illustrated inFigure 1.5. The most common example of a full-duplex network is, once again, the telephone system. Both parties can speak simultaneously during a telephone call and each party can also hear the other at the same time. An example of a half-duplex communication system is a two-way radio like a CB radio, in which only one party can transmit at any one time, and each party must say "over" to signal that he or she has finished talking and is switching from transmit to receive mode.

Figure 1.5   Most LANs use half-duplex communications, meaning that only one side of a connection can transmit at a time

With the right equipment, full-duplex communication is possible on certain types of LANs. The first requirement is a separate channel for traffic running in each direction. Whether this is possible depends on the network medium. Coaxial cable, for example, contains a single conductor and a ground, so there is no physical way that traffic could run in both directions, unless you were to install two cable runs for each connection. Twisted pair cable, on the other hand, contains four separate wire pairs within a single sheath, one of which is dedicated to incoming traffic and one to outgoing traffic. Networks that use this type of cable can therefore theoretically operate in full-duplex mode, and some manufacturers are making Ethernet equipment that makes this possible. Full-duplex Ethernet essentially doubles the throughput of the existing network.

Segments and Backbones

When a small network begins to grow, it is possible to connect LANs together in a haphazard manner for a while. However, building a large enterprise network by connecting many LANs is a complex undertaking that requires careful planning. One of the most common designs for a network of this type is a series of segment LANs connected by a backbone LAN.

The term segment is sometimes used synonymously with LAN or network torefer to any collection of networked computers, but in this context it refers to a LAN composed of user workstations and other end-user devices, such as printers. An enterprise network would consist of many such LANs, all of which are connected to another LAN called a backbone . The backbone exists primarily as a conduit that enables the segments to communicate with each other. One common configuration for an office building with multiple floors calls for a horizontal segment connecting all of the workstations on each floor and a backbone running vertically from the top of the building to the bottom, connecting all of the segments. Such a configuration is illustrated in Figure 1.6.

This type of configuration increases the efficiency of the network by using the backbone to carry all of the traffic going from one network to another. No packet has to traverse more than three LANs using this model. By contrast, if you were to connect each of the horizontal segments to the adjacent segment, daisy chain fashion, most of the internetwork packets would have to travel through many more segments to reach their destinations, burdening the intermediate segments with through traffic for no good reason.

In many cases, the backbone network runs at a higher speed than the segments and may also use a different type of network medium. For example, a typical network might use 10Base-T Ethernet, running at 10 megabits per second (Mbps) over copper UTP cable, for the segments, and it might use 100Base-FX Ethernet, running at 100 Mbps over fiber optic cable, for the backbone. There are two reasons for using a different type of network for the backbone. First, the backbone by definition must carry all of the internetwork traffic generated by all of the segments, and a faster protocol can prevent the backbone from becoming a bottleneck. Second, the backbone may have to span a much longer distance than the segments, and a network that uses fiber optic cable can handle long distances better.

Figure 1.6   An enterprise network spanning an entire building can consist of an individual segment on each floor and a backbone connecting the segments on all of the floors

Client/Server and Peer-to-Peer Networking

Computers can interact with each other on a network in different ways and fulfill different roles. There are two primary networking models used to define thisinteraction, called client/server and peer-to-peer. On a client/server network , certain computers act as servers and others act as clients . A server is simply a computer (or more precisely, an application running on a computer) that provides a service to other computers. The most basic network functions are the sharing of files and the sharing of printers; the machines that do this are called file servers and print servers. There are many other types of servers as well: application servers, e-mail servers, Web servers, database servers, and so on. A client is a computer that avails itself of the services provided by servers.

Although servers are often thought of as computers, they are actuallyapplications. A single computer can conceivably run several different serverapplications at the same time and, in most cases, perform client operations as well.