In the example given in the previous section, the only fault occurs on the link connecting C and D. The router can converge to a new topology and restore access to the network on the Gateway Router D through another path. If D itself fails, it will cause more serious results. The convergence process in the preceding example starts when D notifies B of A link failure. If the link is D rather than C fails, neither B nor C will be able to receive updates, notifying them that the topology has changed.
In this case, convergence to a new topology can lead to a phenomenon called infinite value calculation. When the network becomes completely inaccessible, it is based on the following incorrect idea: there is another router that can access that inaccessible destination. In this case, the router's accounting value r I P is infinite.
To display the inherent risks caused by this catastrophic fault from the routing perspective, reconsider the topology in the convergence diagram. In Figure 1 2-1 2, router D fails.
|
Figure 12-12 router D faults |
Because router D is faulty, all hosts in the network cannot be accessed from outside. Vroc C will invalidate the C-D route and broadcast it as inaccessible after six consecutive updates of vrod D are not received. This is shown in Figure 1 2-1 3. Router A and router B do not know about routing failure until they receive A notification from C.
In this case, A and C believe that B can reach D. They will re-calculate their routes, including this much more expensive roundabout line. Figure 1 2-1 4 shows this.
The two routers send their next update packet to their directly adjacent router B. Router B has timed out its route to D. I believe that A or C can still access D. Obviously, this is impossible, because A and C depend on the link that B just voided. Essentially, A ring is formed between A, B, and C. The formation of this ring is formed by the following incorrect idea: A and C can still reach the router D through the other party. This is because both of them have connections to B, while B has a connection to D.
|
Figure 12-13 router C voided by C-D Routing |
|
Figure 12-14 A and C believe they can access D through B |
During each iteration of an update, the consumed metric increases to the computed Ring Based on the next hop. This type of loop is caused by time delay, and this time delay is characteristic of the independent convergence process of the update packet sent by the adjacent parties.
In theory, the node will eventually realize that D is not reachable. However, it is almost impossible to tell when to converge. This example accurately shows why r I P sets the infinite interpretation to such a small value. Once a network is inaccessible, you must stop this process when you incrementally measure the practical value through updates. This means that the upper bound must be set to a large value. When the calculation reaches the upper bound, a destination cannot be reached. Any upper bound corresponds to the diameter limit of a routing network. In the r I P example, its original designers thought that a 5-hop architecture was already large enough for an autonomous system. A larger system can use more complex routing protocols.