Open Shortest Path First (OSPF) is a link-state routing protocol: every router within an area has a complete map (the Link State Database, LSDB) of all routers and links in that area, and runs Dijkstra's SPF algorithm to compute the shortest path tree. The area hierarchy exists to limit how much of the network each router must hold in its LSDB — and by extension, how much SPF computation it must perform when the topology changes.

Area 0: The Backbone

All OSPF areas must connect to Area 0 (the backbone). Intra-area routing is optimal; inter-area routing is summary-based (Type 3 LSAs from ABRs). All inter-area traffic must transit Area 0 — this is by design, not a limitation. It prevents routing loops in a hierarchical topology.

Area Border Routers (ABRs) sit between Area 0 and non-backbone areas. They maintain two LSDBs (one per area) and generate Type 3 Summary LSAs that advertise the summarised prefixes of one area into the other. This is where route summarisation happens — and where it must happen, since you can only summarise on ABRs in OSPF.

Stub Areas: Reducing External Route Advertisements

A stub area blocks Type 5 LSAs (external routes redistributed into OSPF from other protocols — BGP, EIGRP, static). Instead, the ABR injects a default route (Type 3 LSA) pointing toward Area 0. Routers in the stub area reach external destinations via this default rather than maintaining a full table of external routes.

Use stub areas for:

  • Branch sites or access layer areas that have no local default exit and don't need to know the detail of external routes
  • Areas with limited memory routers (Type 5 tables can be large in environments with many redistributed routes)

Stub area requirements: all routers in the area must be configured as stub (one non-stub router blocks the area from being stub). No ASBR (Autonomous System Boundary Router) can exist inside a stub area, since there are no Type 5 LSAs to distribute.

Totally Stubby Areas: Maximum Summary

An extension to stub areas (Cisco proprietary, but widely supported): block both Type 5 (external) and Type 3 (inter-area summary) LSAs. Only the ABR's default Type 3 LSA remains. Routers in a totally stubby area have the smallest possible LSDB — only intra-area LSAs plus a single default route.

This is the correct design for spoke sites in a hub-and-spoke topology that have a single upstream path and no need for routing granularity beyond "everything goes to the hub."

NSSA: External Routes Without Type 5

Not-So-Stubby Area (NSSA) is the compromise for areas that need to redistribute external routes locally but cannot receive Type 5 LSAs from other areas (because the area is otherwise stub-like). An ASBR in an NSSA generates Type 7 LSAs (instead of Type 5). The ABR translates Type 7 to Type 5 before flooding them into Area 0.

When to use NSSA:

  • A branch site connects to a secondary ISP or a partner network via BGP and needs to redistribute those routes into OSPF — but shouldn't receive the full external route table from other areas
  • A distribution layer area that terminates VPN tunnels (redistributing static routes from VPN peers into OSPF) while remaining shielded from external routes redistributed elsewhere

Virtual Links: When Area 0 Continuity is Broken

OSPF requires all non-zero areas to connect physically to Area 0. In merged networks (post-acquisition) or in network designs where a non-zero area separates two segments of Area 0, a virtual link is a logical extension of Area 0 through a transit area. Virtual links are a workaround, not a design goal — if your network requires virtual links to function, this is a signal that the physical design should be reviewed.

Multi-Area Design for the Enterprise Campus

A typical enterprise campus design:

  • Area 0: Core and distribution layers. ABRs sit at the distribution layer boundary.
  • Area N (access): One area per building or campus zone. Totally stubby if the access layer has no local exit. NSSA if access layer routers have VPN or direct WAN connections to terminate.
  • Summarisation: Each ABR summarises its access area's prefix space into Area 0. Design your IP address plan so each area has a contiguous, summarisable block — this is the single most important planning prerequisite for clean OSPF design.

Tuning for fast convergence: reduce OSPF dead intervals (from default 40s to 10s), enable BFD for sub-second failure detection on point-to-point links, and use OSPF Fast Hello (hello interval 1s) on links where BFD is unavailable.

Sripadatech designs enterprise routing architectures and conducts OSPF optimisation engagements. Contact us to review your IGP design.