Monday 3 October 2016

Distance Vector, Link State, Split horizon and Poison Reverse Explained

Distance Vector routing boils down to two factors - the distance (metric) of the destination and the vector (direction it takes to get there.)

Unlike link-state routing; distance vector only exchanges routing information between it's directly connected neighbors. This means that although a router knows who from whom a router was learnt - it does not know where the neighbor learnt that route from - this means that the router itself will not have a hierarchical view of other subnets and only knows which link / neighbor to use to get to that network.

Examples of routing protocols that employ distance vector routing are RIP and EIGRP (although EIGRP uses Bandwidth, Load, Delay, Reliability and MTU as a metric - AKA K-Values unlike RIP.)

The routing protocols above (by default) employ measures such as split horizon and posion reverse to prevent routing loops.

Link-state routing takes a slightly different approach in that it as aware of the complete OSPF network topology - i.e. it can see a hierarchical view of all of the routers / networks and can see which networks are attached to which routers - but obviously this can become a problem in larger OSPF domains as this might cause performance problems on low spec routers (areas are used to help manage this problem / risk.)

OSPF employs the SPF (shortest path first) algorithm in which it creates a SPF tree with itself as the root node for all known routes.

Link state advertisements (LSAs) are flooded throughout the link-state domain to ensure that all routers in the domain have an up to date copy of area's link state database.

Split horizon is a mechanism that prevents a router from advertising a route back onto the interface it originally received it from - hence preventing a routing loop - routing protocols such as RIP and EIGRP utilize this.

Poison reverse is a mechanism in which a gateway node (a node directly connected to a network) instructs it's directly connected neighbors that the link is down / network is unavailable. In order to perform this the gateway node sets the number of hops to the unavailable network to an 'infinite' amount of hops - hence other nodes in the domain then know that the network is unavailable.


Post a Comment