OSPF concepts: link state, areas and neighbours

Why dynamic routing?

Static routes are fine for five networks. With 50 routers and hundreds of subnets, typing routes by hand is slow, error-prone, and nothing reroutes traffic when a link fails. A dynamic routing protocol lets routers tell each other about networks, and recalculate paths by themselves when something changes. The time it takes all routers to agree on the new paths is called convergence.

Distance vector versus link state

A distance vector protocol, such as RIP, works by rumour: each router tells its neighbours "I can reach these networks at this distance". A router knows only what its neighbours told it, not the real map. RIP uses hop count as its metric, with 15 as the maximum, so it suits only small networks.

A link-state protocol, such as OSPF, works with a map: every router describes its own links (its interfaces, their networks and costs) and floods that description to all routers in the area. Each router collects all the descriptions into the same database and calculates the best paths itself.

LSAs, the LSDB and SPF

• LSA (link-state advertisement): a small message that describes part of the network. A Type 1 (router) LSA is made by every router and lists its links. A Type 2 (network) LSA is made by the DR of an Ethernet segment and lists the routers on that segment.
• LSDB (link-state database): the collection of all LSAs. Every router in the same area has an identical LSDB.
• SPF: the shortest path first algorithm, invented by Edsger Dijkstra. Each router puts itself at the root of a tree and finds the lowest-cost path to every network. The results go into the routing table as O routes.

Areas and the backbone

An area is a group of routers that share the same LSDB. Big networks are split into areas so that each router keeps a smaller map and a link flapping in one area does not make every router recalculate. Area 0 is the backbone: every other area must connect to it. A router with interfaces in area 0 and another area is an ABR (area border router). For the CCNA you mainly configure single-area OSPF, with everything in area 0.

Router ID

Every OSPF router needs a router ID (RID): a 32-bit number written like an IPv4 address, unique in the OSPF domain. It is chosen in this order:

1. The router-id command, if configured (best practice).
2. Otherwise, the highest IP address on an up loopback interface.
3. Otherwise, the highest IP address on an active (up/up) physical interface.

The RID is chosen when the OSPF process starts. If you change it later, it takes effect only after clear ip ospf process or a reload.

Cost: the OSPF metric

Cost = reference bandwidth / interface bandwidth, rounded down, with a minimum of 1. The default reference bandwidth is 100 Mbps. The cost of a path is the sum of the costs of the outgoing interfaces along the way.

With the default, every link of 100 Mbps or more costs 1, so OSPF cannot tell a 100 Mbps link from a 10 Gbps one. Fix it with auto-cost reference-bandwidth 100000 under router ospf (the value is in Mbps). Set the same value on every router, or paths will be calculated inconsistently.

Hellos and neighbour states

OSPF routers find each other by sending hello packets to multicast 224.0.0.5 (OSPF runs directly over IP as protocol 89). On broadcast (Ethernet) and point-to-point links, hellos go every 10 seconds and the dead interval is 40 seconds: if no hello arrives for 40 seconds, the neighbour is declared down. Two routers become neighbours only if key settings match, such as area, subnet and timers.

DR and BDR on Ethernet

On a shared Ethernet segment with many routers, if every router became fully adjacent with every other, the flooding would multiply. Instead, the routers elect a DR (designated router) and a BDR (backup DR). Every router becomes Full only with the DR and BDR; the others (DROTHERs) stay in 2-Way with each other, and that is normal. The highest OSPF interface priority wins (default 1, range 0 to 255); a tie goes to the highest RID. Priority 0 means the router never becomes DR or BDR. The election does not preempt: a new router with a higher priority does not take over from a working DR.