Why do networks need OSPF?

A real-life situation

A company has its office network, LAN A, behind router R1, and its servers, LAN B, behind R4. Two routers sit in between, R2 and R3, so there are two ways from the office to the servers: over the top through R2, or along the bottom through R3. The second path was bought for one reason: if something on the first path breaks, traffic should keep flowing.

An engineer set the network up with static routes, all pointing over the top. One evening the cable between R2 and R4 is cut. The bottom path is perfectly healthy, but the office can't reach a single server until someone logs in and changes the routes by hand. This lesson explains why, and what OSPF does differently.

The network in this module

The three lessons in this module use the same four routers. Each router-to-router link is its own small subnet (a /30), and every router has a router ID (RID), which the next lessons explain.

Where a router's routes come from

A router forwards a packet only if its routing table has a route that matches the destination; otherwise it drops the packet. Routes get into the table in three ways:

• Connected routes are added automatically for every subnet the router has an interface in. R1 knows LAN A (192.168.1.0/24) this way, but nothing about LAN B.
• Static routes are typed in by an engineer, one line per destination.
• Dynamic routes are learned from other routers through a routing protocol. R4 tells its neighbours "I am connected to 192.168.4.0/24", they pass that on, and R1 ends up with a route to LAN B that nobody typed.

Because R1 hears about LAN B over two paths, a routing protocol also has to choose: each protocol has a metric, a number that says how good a path is, and the path with the best metric goes into the routing table.

What static routes can and can't see

A static route is a line of configuration: "to reach 192.168.4.0/24, send packets to 10.0.12.2". On R1 it looks like this:

The router keeps that route in its routing table for as long as it can reach the next hop through one of its own interfaces. In simple terms, a static route only notices a failure on the router's own link. If R1's Gi0/1 goes down, the route to 10.0.12.2 disappears and the static route is removed. If the failure is one hop further away, between R2 and R4, R1's interface stays up, 10.0.12.2 still answers, and R1 keeps sending everything to R2.

A floating static route (a backup route with a higher administrative distance) doesn't help here either. It only takes over when the main route leaves the routing table, and R1's main route is still there.

Static routes at scale

Even without failures, static routes become a lot of typing. Every router needs a route to every network it is not directly connected to. The four-router network above has six networks: two LANs and four router-to-router links.

Fourteen routes for four routers is manageable. A network of 40 routers and 200 subnets needs thousands, and every new subnet means visiting every router. Each one is a chance for a typing mistake, and a mistake in a static route usually shows up as a routing loop or a black hole that is hard to trace.

What a dynamic routing protocol does instead

With a routing protocol such as OSPF, the routers do the work themselves. They:

• find their neighbours: other routers on the same links running the same protocol;
• share what they know: which networks they are connected to and how good each link is;
• pick the best path to every network using a metric (OSPF uses cost);
• notice failures anywhere in the network, not only on their own links;
• recalculate and switch to another path without anyone logging in;
• learn new networks as soon as they are added on any router.

The time from a change (a link failing, a router rebooting) until every router agrees again on the best paths is called convergence. Fast convergence is the main reason networks run a routing protocol at all.

Why it works this way

A static route is a fixed instruction about the next hop only. The router has no information about the rest of the path, so it can't know when a link two hops away has failed. A routing protocol exists to give routers that missing information: each router tells the others about its links, so a failure anywhere is reported everywhere.

That information costs something: CPU and memory to run the protocol, and a little bandwidth for the messages. That trade-off is why many networks use both. OSPF runs between the core routers, and a single static default route points out to the ISP, where there is only one way out anyway.

Static or dynamic?

Two ways to group routing protocols

By where they run. An autonomous system (AS) is a network under one organisation's control. An interior gateway protocol (IGP) such as OSPF, EIGRP, IS-IS or RIP shares routes inside one AS. An exterior gateway protocol (EGP) shares routes between different organisations; today that is only BGP. ("Gateway" is the old word for a router.)

By how they learn. This is the difference that explains how OSPF behaves:

A simple way to picture the difference: a distance-vector router follows signposts ("LAN B, 3 hops, that way"), while a link-state router carries a map and plans the route itself. EIGRP sits in between. It shares routes with neighbours like a distance-vector protocol but adds features such as backup routes kept ready in advance, so it is often called an advanced distance-vector or hybrid protocol.

Why OSPF and not another protocol?

The CCNA compares these protocols briefly; OSPF is the one you configure.

• RIP counts hops only, so a slow link with one hop beats a fast link with two. It allows at most 15 hops and converges slowly. It is mostly found in labs and very old networks.
• EIGRP converges fast and was developed by Cisco. Its basic form was published in 2016 (RFC 7868), but in practice it runs almost only on Cisco equipment.
• OSPF is an open standard (RFC 2328 for IPv4), so routers and firewalls from different vendors can share routes. It converges quickly, chooses paths by link speed, and scales to large networks by splitting them into areas.
• BGP connects separate organisations across the internet. It is covered in Where BGP fits.

💡 Exam tip: expect questions that ask what happens when a link beyond the next hop fails on a statically routed network. The answer is that the static route stays in the table and traffic is dropped further along. Also know that a floating static route needs a higher administrative distance than the route it backs up.

Common mistakes

• Assuming a static route is removed when any link on the path fails. Only a failure of the router's own exit stops it.
• Expecting a floating static route to cover a remote failure. It only replaces a route that has left the table.
• Thinking dynamic routing replaces static routes entirely. A static default route to the ISP is still normal.
• Calling OSPF a Cisco protocol. It is an open IETF standard; EIGRP is the one that came from Cisco.

Key takeaways

Check yourself