Redundant Routing Protocols

One way to work around incompatibilities in routing protocols is to implement them redundantly. This approach can be implemented in varying degrees, depending on your needs. On one extreme, for example, you could implement two routing protocols on every router in your internetwork. This would

Internetworking with Dissimilar Protocols result in each router

• Having two sets of routing tables

• Performing two sets of table updates

• Converging twice after topology changes

In other words, each router would have twice the work to do in addition to the usual responsibility of forwarding datagrams on behalf of end systems. Therefore, using redundant routing protocols is unavoidably resource intensive! The router's internal resources, including CPU cycles and memory, are consumed at prodigious rates. The network can also suffer as bandwidth consumption on the network increases. The extent to which each of these resources will be taxed depends directly on

• The size of the internetwork using redundant protocols

• The protocols used

• The actual metrics or settings implemented

Given these variables, it may not be easy or feasible to assess the actual effects of using redundant routing protocols. Nevertheless, these factors adversely affect the network's performance. Unlike with routed protocols, however, it doesn't make sense to use fully redundant routing protocols. Each protocol would be capable of calculating routes through the network to any given destination, so there is nothing to be gained from full routing protocol redundancy.

A more sensible approach is to partition off just the pieces of the internetwork that require redundant protocols and use a single protocol everywhere else. This would improve the operational efficiency of the routers, as well as the overall network. You may elect, for example, to create stub networks within your internetwork. Each stub could use a simple, distance-vector protocol such as RIP or RIP-2. The remainder of the network would use a more robust protocol such as OSPF or EIGRP. In such a scenario, only the border routers between the two network regions would have to support both protocols. Figure 14-6 illustrates this.

Figure 14-6: Using different routing protocols for different network regions.

Figure 14-6: Using different routing protocols for different network regions.

In the network depicted in Figure 14-6, the use of different routing protocols for different parts of the network improves overall network performance. Specifically, topological change wouldn't trigger a networkwide convergence. Each network region need only track the topology of its own region. Therefore, the change were to occur within the OSPF region, only the OSPF routers would have to converge on a new understanding of the network's topology. Similarly, if the change were to occur within the RIP region, the OSPF routers wouldn't need to know about the new topology.

The OSPF routers would, however, need to know whether the change meant that entries in their routing tables were now invalid. This information would be gained via the exchange of routing information with RIP nodes. Although it is quite normal for routers with the same routing protocol to exchange routing information, a logistical challenge arises when two neighboring routers actually use different routing protocols. The term that describes the exchange of routing information between dissimilar routing protocol nodes is route redistribution. In the example presented in Figure 14-6, the border router(s) must redistribute routing information from the RIP network into the OSPF network and vice versa.

0 0

Post a comment