Routers

Routers

Bridges are suitable for relatively simple networks, but bridges have certain limitations that become more significant in complex network situations. One limitation of bridges is that a network with bridges generally cannot include redundant paths. (Redundant paths are desirable because they enable the network to continue functioning when one path goes down.)

Consider the network in Figure 6.8. Both bridges are aware of the existence of Node B, and both can pick up the packet from Net A and forward it. At the very least, the same packet can arrive twice at Node B.

A worse case, however, is that these relatively unintelligent bridges can start passing packets around in loops, which results in an ever-increasing number of packets that circulate on the network and never reach their destinations. Ultimately, such activity can (and will) saturate the network.

An algorithm, called the spanning tree algorithm, enables complex Ethernet networks to use bridges while redundant routes exist. The algorithm enables the bridges to communicate and construct a logical network without redundant paths. The logical network is reconfigured if one of the paths fails.

Another problem is that the bridges cannot analyze the network to determine the fastest route over which to forward a packet. When multiple routes exist, this is a desirable capability, particularly in wide area networks (WANs), where some routes are often considerably slower than others.

Routers organize the large network in terms of logical network segments. Each network segment is assigned an address so that every packet has both a destination network address and a destination device address.

Recall that an internetwork consists of two or more logically separate but physically connected networks. By this definition, any network segmented with routers is an internetwork.

Routing in Windows NT

The word “router” evokes the image of a screenless, box-shaped device—and many routers fit that image—but the tasks performed by a router can be (and sometimes are) performed by a PC. Many situations exist in which it is useful to configure a PC for routing functions, and one of the most important is when the PC serves as a remote access server. Windows NT 4.0 Remote Access Service (RAS), for example, is capable of acting as an IP or an IPX router or a NetBIOS Gateway. (More on the NetBIOS Gateway in the next sidebar.) Modem connections used to enable only point-to-point communications (from one computer to another computer). Under Windows NT, the connection is still a point-to-point connection (from the remote client computer to the RAS server on the local LAN), but the RAS server can route packets to other computers, thus providing the remote client with access to the entire network (see Figure 6.9).

Routers are more “intelligent” than bridges. Not only do routers build tables of network locations, but they also use algorithms to determine the most efficient path for sending a packet to any given network. Even if a particular network segment isn’t directly attached to the router, the router knows the best way to send a packet to a device on that network. In Figure 6.10, for example, Router A knows that the most efficient step is to send the packet to Router C, not Router B.

Notice that Router B presents a redundant path to the path Router A provides. Routers can cope with this situation because they exchange routing information to ensure that packet loops don’t occur. In Figure 6.10, if Router A fails, Router B provides a backup message path, thus making this network more robust.

One consequence of all the processing a router performs on a packet is that routers generally are slower than bridges.

You can use routers to divide large, busy LANs into smaller segments, much as you can use bridges. But that’s not the only reason to select a router. Routers also can connect different network types. Notice that the network in Figure 6.10 includes a Token Ring segment with the Ethernet segments. On such networks, a router is the device of choice.

The protocols used to send data through a router must be specifically designed to support routing functions. IP, IPX, and DDP (the AppleTalk Network-layer protocol) are routable protocols. NetBEUI is a nonroutable protocol.

Because routers can determine route efficiencies, they usually are employed to connect a LAN to a wide area network (WAN). WANs frequently are designed with multiple paths, and routers can ensure that the various paths are used most efficiently.

The Network layer functions independently of the physical cabling system and the cabling system protocols—independently, that is, of the Physical and Data Link layers. This is the reason that routers easily can translate packets between different cabling systems. Bridges, on the other hand, cannot translate packets in this way because they function at the Data Link layer, which is closely tied to physical specifications.

Routers come in two types:

Static Routers. These routers do not determine paths. Instead, you must configure the routing table, specifying potential routes for packets.
Dynamic Routers. These routers have the capability to determine routes (and to find the optimum path among redundant routes) based on packet information and information obtained from other routers.

To determine the best path for a packet, routers employ some form of routing algorithm. Some common routing algorithms are discussed in the following sections.

Further Information

Routing Algorithms