Figure 815 A and C believe that they can access D through B

Nelifcui * AutortDrY /jus aysleni.

Nelifcui * AutortDrY /jus aysleni.

192."cS.1£5.1D -92.1«. 125.15

These routers send their next updates to B, an immediate neighbor of both routers. Router B, having timed out its own route to D, believes it can still access D through either A or C. Obviously, it cannot because those routers are relying on the link that B just invalidated. In essence, a loop is formed between A, B, and C that is fed by the mistaken belief that both A and C can still reach the unreachable Router D through each other. This is because both have a connection to B, which has the connection to D.

With each iteration of updates, the cost metrics are incremented to account for the next extra hop that is added to the loop already calculated. This form of looping is induced by the time delay that characterizes independent convergence through neighbor- transmitted updates.

In theory, the nodes will eventually realize that D is unreachable. It is virtually impossible, however, to tell how much time would be required to achieve this convergence. This example illustrates precisely why RIP's interpretation of infinity is set so low! Whenever a network becomes inaccessible, the incrementing of metrics through routine updates must be halted as soon as practical. Unfortunately, this means placing an upper limit on how high the nodes will count before declaring a destination unreachable. Any upper limit directly translates into a limitation on the maximum size of the routed network's diameter. In the case of RIP, its original designers felt that 15 hops were more than adequate for an autonomous system. Systems larger than this could utilize a more sophisticated routing protocol.

RIP supports two means of avoiding the count to infinity loop problem:

• Split horizon with poisoned reverse

• Triggered updates

0 0

Post a comment