Divide a OSPF Network into Areas
As an OSPF network gets larger, one of the challenges is keeping all the link-state statements on all the routers in sync. One way to control the size of the OSPF network is to divide it into smaller pieces, which OSPF calls areas.
Each area has the same properties: All the routers within the area exchange their network topology information in LSAs, and this smaller group of routers runs the SPF calculation to keep its link-state databases identical. An OSPF network must have at least one area, the backbone area, called Area 0.
Routers that connect only to other routers within an area are called internal routers.
Some of the routers within an OSPF area sit on the boundary of two areas. These area border routers (ABRs) connect the areas within the larger OSPF network. They run two SPF calculations, one for each area to which they’re connected, and they maintain two link-state databases, one for each area.
The ABRs pass route information between the two areas, but condense (summarize) it before sending it into the neighboring area. The summarization improves the overall stability of the OSPF network.
Areas and area border routers sit in a single OSPF administrative domain. Another type of router sits at the boundary between one OSPF administrative domain and another administrative domain or routing protocol. This router is the AS boundary router (ASBR). These routers are responsible for advertising externally learned routes into the OSPF administrative domains.
Areas come in several different flavors. The main one is the backbone area because it forms the backbone of the OSPF network. The backbone area has the area ID 0, which is normally written as the 32-bit value 0.0.0.0.
All areas in the OSPF network connect to the backbone, and as a result, all area border routers are part of the backbone area. Also, any networks that have an area ID of 0.0.0.0 must also connect to the backbone area.
All routers in the OSPF backbone must be physically connected to each other. If any routers aren’t physically contiguous, they must be connected by an OSPF virtual link so that they appear to be contiguous.
In a simple (if you can conceive of that right now) OSPF network, all non-backbone areas (areas numbered other than 0) connect directly to the backbone area (area 0). By default, these non-0-numbered areas are referred to as regular areas.
To minimize the amount of LSA traffic on segments of the OSPF network, OSPF has two other types of areas that don’t advertise route information into other areas. Stub areas receive only summarized routing information about other areas within the OSPF domain, and they don’t receive any information about external OSPF routes. As a result, stub areas can’t connect to external networks.
Not-so-stubby areas (NSSAs) are a slight variant of a stub network. These areas can connect to external networks.