Distance vector (e.g., RIP)
- What is advertised is the path (vector) and its cost (distance, typically just the number of hops), hence, the name of this routing protocol category.
- periodic flooding of routing tables (every 10 - 90s).
- Thus, long convergence time and useless overhead in case of stable network
- Adapted to only small networks. That's why RIP is limited to networks with less than 15 routers.
Link state (e.g., OSPF, IS-IS)
- What is advertised is the link state (all links and their cost), hence, the name of this routing protocol category.
- Floods only topology changes (link state advertisements): faster convergence time and less overhead than distance vector protocols. Uses keep-alive messages to keep connection to other routers.
- All routers recalculate routes when topology changes: requires CPU and memory, need to partition the network into areas to avoid having 100 routers recalculating their routing table for every link state change (imagine the case of a flapping link...).
Path vector (e.g., BGP)
- What is advertised is the path (vector), without distance. The decisions are made based on other elements (typically predefined policies) than a path cost.
|The routing complexity of setting up VPNs accross the Internet|