Open Shortest Path First Protocol

Ron McCarty

In the March 2001 issue of Sys Admin, I covered the Routing Information Protocol (RIP). As discussed in that column, RIP is a distance-based protocol where each node advertises its complete routing table every 30 seconds. The distance is based on a maximum hop count of 15, where 16 represents infinity or a poisoned route that is to be dropped. The limited hop count, along with the practice of advertising the complete routing table, prevents RIP from being the protocol of choice for large networks.

Opens Shortest Path First (OSPF), on the other hand, is a link state protocol. A link state protocol advertises changes within the network, not the complete routing table. For example, consider the network shown in Figure 1. Using a link state protocol like OSPF, Router A tells Routers B and C that link A1 has failed. The state changes can immediately be sent when Router A notices that A1 has failed, for example, when the CRC errors become so high as to make the link unusable or when it loses its electronic signal. If RIP or another distance vector protocol were used, Router A would wait until its next update (in 30 seconds) to send out its routing table. The routing table would not include the route for path A1, and the change would “slowly” make its way through the network or “converge”. OSPF’s current version, version 2, is defined in RFC 2328, “OSPF Version 2”.

Each OSPF node or router uses a link state advertisement (LSA) to tell other routers about link states it has received from other routers or neighbors through other LSAs and its own links to which it is directly connected. Because only changes are advertised, OSPF routers — in addition to recognizing failures caused by physical or lower-level protocols — must also have a mechanism for determining when an OSPF neighbor has failed.