Suppose you have two or more individual LAN segments and the goal is to bring them together.  It is relatively easy if these two segments are located on the same floor and use the same network architecture, for example 10Base2.  In this case, connecting these separate segments together can be as simple as using a BNC barrel connector.

What if these segments use different network architecture?  Let’s say, for example, one of them is an Ethernet network and another is Token Ring.  In this case, simply connecting the segments together will not make the system work because of the different media access control methods used in the respective network architectures.  The computers will not understand each other and will not be able to decide who transmits and who keeps silent.  Even if the networks were all of the Ethernet type but some of them operated at 100 Mbps and others at 10 Mbps, the speed difference would be still make it impossible to simply connect the wires together.  We will need to use some intelligent devices to handle the differences in the media access control in the first example and the speed difference in the second.

The situation becomes even more complex when the LAN segments are distributed across distances greater than the maximum distance allowed by the local area network technology.  For example, if two Ethernet networks are located in different cities, we cannot just use a single coaxial cable to connect them because of attenuation considerations.  Someone could propose using repeaters to overcome the signal attenuation. This solution, however, will not work because of the propagation delay limitation discussed in Chapter 4 that limits the maximum number of repeaters in Ethernet-type networks.  Similar restrictions exist for Token Ring, ArcNet, 100VG-AnyLAN, and other local area network architectures.  You’ll need to use mechanisms such as point-to-point links or circuit switched solutions to connect these remote locations. We’ll discuss these mechanisms a bit later in the chapter.