There is split-horizon, poison-reverse, hold-down, triggered update mechanisms to handle count to infinity problem in RIP. I am a bit confused about why so many mechanisms are used?when and in what practical scenarios is each applicable? Is there a single mechanism which takes care of count to infinity problem. For example RIP RFC says that triggered updates need to be used when there are more than 3 gateways in a routing loop. How does one know in advance which mechanism to use?
Also, I read some documentation which says that OSPF converges quickly as compared to RIP bcos it is link state based. What exactly does this mean? How is a OSPF routing update different from RIP update ?
The best description of these routing protocols that I have seen is in Routing TCP/IP, volume 1, second edition Jefff Doyle
It is widely respected and I have it right here:)
I'll give Link State a quick go.
RIP as you know advertises routes out of each interface. Each router frequently advertises all of the routes it has learned about, perhaps with regard to split horizon.
OSPF (and ISIS) Do not advertise routes. Ther follows a very brief and idealised description of the principle of link state routing. Each router advertises itself and all of the networks to which it is connected. In this way all routers can construct a full network diagram. This is called the Topology Database. Each router then independently works out which next hop to use for each destination network. The best of these is used to construct the routing table for that router. Once the topology database is constructed no further updates are sent (sort of:).
OSPF does not therefore need (by principle) the distance vector tricks and optimisations to control routing loops and to speed convergance times.
It does however have its own set of tricks and optimisations which are required to control the volume of updates needed. Multiple Areas, Designated Routers, Stub areas. Also the whole topology database is updated periodically.
There will be for sure some decent descriptions on-line but I cant point you there since I have Doyle already:)