Solved: Split horizon

vijayendran · ‎05-14-2003

Is split horizon applicable only to Distance Vector Routing Protocols ?

If Split Horizon is not applicable to Link State Routing Protocols, how are routing loops avoided in Link State Routing Protocols ?

mark-obrien · ‎05-14-2003

Split Horizon only applies to distance vector protocols. Routing loops are avoided in link state protocols because these protocols are designed with the assumption that a router will receive the same LSA, to use OSPF's nomencature - other protocols have different names, from several different sources. When a router receives an LSA, it compares the sequence number of this LSA with the sequence number of the same LSA in its database, if it exists. If the newly received LSA has a higher sequence number than the instance of the LSA in the database, the router replaces the LSA in its database with this newly received instance of the LSA.

Plus, since link state protocols advertise links instead of routes, there is no risk of a router receiving an LSA from another router and not know whether or not this is an LSA that the router itself originated. Compare this to a distance vector protocol. A router will send a route to a neighbor, and without split horizon that neighbor will send the route back, and the first router will not know whether this is a valid route, or just the same route that it just sent.

I hope this clears things up.

Mark

View solution in original post

mark-obrien · ‎05-14-2003

Split Horizon only applies to distance vector protocols. Routing loops are avoided in link state protocols because these protocols are designed with the assumption that a router will receive the same LSA, to use OSPF's nomencature - other protocols have different names, from several different sources. When a router receives an LSA, it compares the sequence number of this LSA with the sequence number of the same LSA in its database, if it exists. If the newly received LSA has a higher sequence number than the instance of the LSA in the database, the router replaces the LSA in its database with this newly received instance of the LSA.

Plus, since link state protocols advertise links instead of routes, there is no risk of a router receiving an LSA from another router and not know whether or not this is an LSA that the router itself originated. Compare this to a distance vector protocol. A router will send a route to a neighbor, and without split horizon that neighbor will send the route back, and the first router will not know whether this is a valid route, or just the same route that it just sent.

I hope this clears things up.

Mark

godlam · ‎05-14-2003

How about EIGRP? I think OSPF and EIGRP is not applied the split horizon.

mark-obrien · ‎05-15-2003

OSPF does not run split horizon, but EIGRP does by default. See this link:

http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122cgcr/fiprrp_r/1rfeigrp.htm#1018037

EIGRP has additional methods of verifying loop-free routes. A router has a metric for a given route, and it holds this metric as a "feasible distance". Any neighbor advertising a metric less than the feasible distance for that route is listed as a "feasible successor", or a potential next-hop in the event of a failure of the route. By definition, if a router sees the route advertised by a neighbor with a metric less than the feasible distance, the neighbor must have learned of the route by independent means, and not from the first router.