When all routers in a internetwork are operating with the same knowledge, the internetwork is said to be converged. That is, all routers in your internetwork use a routing protocol (such as RIP) and send their routing table to the routers connected to them. Any changes in the routing table are updated at a certain interval. For RIP its every 30 seconds, for IGRP its every 90 seconds.
Fast convergence is significant because it reduces the period of time in which routers would continue make incorrect routing decisions based on incorrect routing tables. These incorrect routing tables can happen due to interfaces being shut down, IP addresses being changed, etc. A fast convergence allows your network to run smoothly without your routers sending data to the wrong destinations.
Re: "convergence time" in routing protocols.
15 years 5 days ago #1702
Rabbit's given you a very good answer. You can think of convergence as the amount of time it takes the whole network to agree on the best paths to all networks. One of the main drawbacks of RIP as a protocol is the fact that it has a very slow convergence time.
One of the drawbacks, as Sahirh mentions, if RIP is the convergence time is slower than IGRP or EIGRP. The problem with slow convergences is the occurrence of routing loops, which can render you subnetworks useless.
The different protocols use different methods to prevent these loops from happening: holddown timers, poison reverse updates, split horizons. If the convergence is slower than these method, you can still run into your routing loops.
Sometimes, in the case of IGRP, you can turn off your timers to speed up the convergence, but then you run the risk of these routing loops.