OSPF

networking

Open Shortest Path First: the standard link-state interior routing protocol, in which every router maps the whole area and computes shortest paths.

Runs Dijkstra over a shared link-state database; convergence is fast because everyone recalculates from the same map.

OSPF is the interior routing protocol most enterprises actually run. Every router in an area floods link-state advertisements until all of them hold an identical database, then each independently runs Dijkstra to compute shortest paths. Because the inputs are identical, the results agree, which is what makes convergence predictable.

Areas exist to keep that computation bounded. Flooding everything everywhere does not scale, so the topology is divided, with area zero as the backbone every other area must touch. Border routers summarize between areas, which reduces both database size and the blast radius of a flapping link, and the discipline of designing areas properly is most of what separates a stable OSPF network from a fragile one.

The costs are worth stating honestly. Cost is derived from bandwidth by a reference value that has not kept up with modern interface speeds, so anything above the reference ties unless you adjust it. Adjacencies depend on matching timers, area IDs, authentication and MTU, and an MTU mismatch produces the memorable failure where neighbours reach the exchange state and stick there forever.

← All glossary entries