OSPF areas and why the design matters

OSPF areas are often treated as a naming convention. They are not. An area is a boundary for the link-state database, and that boundary determines how much every router inside it must compute.

What an area actually contains

Every router in an area holds an identical database describing every link in that area, and independently runs Dijkstra over it. Add a link, and every router in the area recalculates. Flap a link, and they all recalculate again.

In a single-area design, that means every router — including the small one at a remote site — carries the full topology of the whole network and recomputes on every change anywhere in it.

What areas buy you

An area border router summarises. Routers in area 10 learn that area 20’s prefixes are reachable through the border router, without holding area 20’s topology. A flap inside area 20 does not cause a recalculation in area 10.

That is the point: areas limit the blast radius of topology change, and they limit memory and CPU on the smallest devices.

The rules that constrain the design

Every area must touch area 0, the backbone, either directly or through a virtual link. This is not arbitrary: it prevents routing loops between areas, because inter-area traffic always transits the backbone.

Stub areas

If an area has one exit, its routers do not need external route detail. A stub area replaces those with a default route. A totally stubby area goes further and replaces inter-area routes as well. A remote site with one uplink might hold three routes instead of three hundred — and recalculate almost never.

area 20 stub no-summary
show ip ospf database database-summary
show ip ospf statistics       # how often is SPF running?

When to bother

Below roughly fifty routers, a single area is usually fine. Beyond that, or where links are unstable, or where remote devices are small, areas stop being tidiness and start being the thing keeping your network converged.