## Convergence

The most significant implication of a topology change in a RIP internetwork is that it changes the solution set of neighboring nodes. This change may also result in different results the next time the distance vectors are calculated. Therefore, the new sets of neighboring nodes must then converge, from different starting points, on a consensus of what the new topology looks like. This process of developing a consensual perspective of the topology is known as convergence. In simple terms, the routers develop an agreement of what the network looks like separately, together.

Figure 8-10 illustrates convergence; it demonstrates the following four possible routes to Router D:

• From Router A, to Router B, to Router C, to Router D

• From Router A, to Router C, to Router B, to Router D

Vendor-specific implementations of RIP that can remember multiple routes can use these four possible routes to reach Router D. Such route diversity affords protection from failures in the network. Of these four routes, the two that use the link between Routers B and D are the least desirable. This is because that link has a metric cost of 10, compared to a cost of 1 for the other routes.

Implementations of RIP that adhere rigidly to RFC 1058 will only remember one of these routes. Ostensibly, this will be the least-cost route that it learns about. For the sake of this example, the primary route to Router D's network is via Router C. If this route were to fail, it would take some time for all the routers to converge on a new topology that didn't include the link between Routers C and D.

As soon as the C-D link fails, it is no longer usable, but it may take a little time for this fact to become known throughout the network. The first step in convergence is for D to realize that the link to C has failed. This assumes that Router D's update timer elapses before Router C's timer. As this link was the one that should have carried updates from Router D to Router C, no updates can be received. Consequently, C (as well as A and B) is still unaware that the C-D link has failed. All routers in the internetwork will continue to forward datagrams addressed to Router D's network number, through that link. Figure 8-11 illustrates this first stage in convergence.

Figure 8-10: Two possible paths to Router D.

Figure 8-10: Two possible paths to Router D.

Figure 8-11: Only Router D is aware of the link failure.

Upon expiration of its update timer, Router D will attempt to notify its neighbors of its perception of the change in the network's topology. The only immediate neighbor that it will be able to contact is B. Upon receiving this update, B will update its routing table to set the route from B-C to infinity for all destination addresses in the 172.31 network, which is the network connected via Router D. This will allow it to resume communications with D, albeit via the B-D link. After B has updated its table, it can advertise its newfound perception of the topology to its other neighbors, A and C.

Note Remember that a RIP node invalidates a route by setting its metric to 16, the RIP equivalent of infinity.

As soon as A and C have received updates, and have recalculated network costs, they can replace their obsolete entries that used the C-D link with the B-D link. The B-D route was previously rejected by all nodes, including B, as being more expensive than the C-D link. Its cost metric of 10 compared unfavorably with the C-D cost of 1 for each node. Now, with the failure of the C-D link, the B-D link features the lowest cost. Therefore, this new route replaces the timed-out route in the neighbors routing tables.

When all routers agree that the most efficient route to D is via B, they have converged. Figure 8-12 illustrates this.

Figure 8-12: The routers converge on B-D as the new route.

Figure 8-12: The routers converge on B-D as the new route.

The amount of time that will elapse before convergence completes is not easy to determine. It will vary greatly from network to network, based on a wide variety of factors that include the robustness of the routers and transmission facilities, amount of traffic, and so on.

0 0