Why do networks need OSPF?

Why does a router need a route before it can forward?

PC-A is a user's computer on LAN A. It needs to open an internal website on a server at 192.168.4.10 on LAN B. The server's address is an IP address, not a website URL. The computer sends packets towards that address through its local router, R1.

R1 checks its routing table: a list of networks and how to reach them. A matching route tells R1 the next router to send the packet to. This router is the next hop. R1 cannot use a path just because a cable exists; it needs a route that chooses that path.

Where does that route come from?

OSPF (Open Shortest Path First) is a routing protocol: rules that let routers share network information and calculate routes. It is one way to supply the remote routes R1 needs.

• Connected route: R1 knows about a network on one of its own working, addressed interfaces.
• Static route: an administrator enters an instruction telling R1 how to reach another network.
• Dynamic route: a routing protocol, such as OSPF, learns network information and calculates a path.

These sources can exist on the same router. This lesson compares static routes with OSPF so you can see when automatic updates help.

Two paths should mean a way around a failure

R4 is the router connected to the server's LAN. R1 can reach it through R2 or through R3. R2 and R3 are routers along the way, not servers or switches. The two paths provide an alternative if one router-to-router connection fails.

OSPF gives each connection a cost, a number used to compare paths. It adds the outgoing costs along each path to the destination network. In this example, the upper path costs 3 and the lower path costs 12, including the last connection to LAN B. For these routes to the same network, OSPF chooses the lower total.

OSPF builds a map, then calculates a route

R4 creates records describing its network connections. These are called link-state advertisements (LSAs). The routers pass these records to other routers in the OSPF area. An area is a part of the OSPF network that shares detailed link information.

Each router stores the records and calculates paths from its own position. R1 learns that LAN B is behind R4 and chooses the path through R2. R4 also needs a route back to LAN A so the server's replies can return.

Now try the same network with static routes

Instead of running OSPF, configure R1 to send LAN B traffic to R2. The other routers also need forwarding and return routes. This command is entered in R1's global configuration mode:

The failure is beyond R1's next hop. R1 can still resolve 10.0.12.2 over its connected link, so the static route remains. Assume R2 has no alternate or default route to LAN B: it drops the traffic. A floating static route on R1 cannot take over while that primary route stays installed. Tracking reachability can change this behaviour; the example has no tracking.

With OSPF, the failure changes the map

1. R2 and R4 detect that their connection has failed.
2. They report the change using updated OSPF records.
3. R1 receives the information and calculates its paths again.
4. R1 replaces the path through R2 with the working path through R3.

This process is called convergence: the routers process a change and settle on updated routes. It takes time, so packets may be lost while the change is being processed.

The new route costs 12. Valid return routes and permitting security policies are still required; learning a forward route alone does not prove that an application will work.

More subnets mean more routes to maintain

With static routing, you must update the relevant routers when you add a remote subnet or change a path. A subnet is an IP network such as LAN B's 192.168.4.0/24. More routers and subnets can mean more entries to check.

In this example, there are six subnets. R2 connects directly to two of them. If you use one static route for each remote subnet, with no default or summary routes, R2 needs four static routes. The other routers also need the correct routes, including paths for replies.

A static default route can be a sensible choice for a branch with one exit. OSPF helps when routers must learn changing networks and choose among paths, at the cost of protocol processing, memory and control traffic. The two approaches can coexist.

Try it: watch a route change

Use an isolated lab with the diagram's four routers already configured for OSPF. On R1 at the privileged EXEC # prompt, run show ip route 192.168.4.10. The route should use R2 at 10.0.12.2 and show cost 3 for this example.

Disconnect only the R2–R4 lab link. After OSPF processes the change, repeat the command: the next hop should be R3 at 10.0.13.2, with cost 12. Reconnect the link and confirm the lower-cost route returns. This test assumes the illustrated costs and no preferred competing route.

Next, see how OSPF works. The protocol specification is RFC 2328.

Test yourself on static and dynamic routing