Yes, OSPF is a link-state routing protocol. It is formally classified this way in RFC 2328, the Internet standard that defines OSPF Version 2: “The OSPF protocol is based on link-state or SPF technology.” This distinction matters because it fundamentally shapes how OSPF learns about a network, how fast it reacts to changes, and how it differs from older protocols like RIP.
What “Link-State” Actually Means
In a link-state protocol, every router builds a complete map of the network rather than relying on secondhand information from neighbors. Each router maintains its own database describing the full topology of the network it belongs to. When a link goes up, goes down, or changes cost, the affected router floods that update to every other router in its area. Every router then holds the exact same topology database and independently calculates the best path to every destination.
This is the opposite of how distance-vector protocols like RIP work. A distance-vector router only knows what its immediate neighbors tell it: “I can reach network X, and it’s Y hops away.” It never sees the full picture. Networking engineers sometimes call this “routing by rumor,” because each router is trusting secondhand (or third- or fourth-hand) information. A link-state router, by contrast, has the raw topology data and draws its own conclusions.
How OSPF Builds Its Map
OSPF routers share topology information through messages called Link State Advertisements, or LSAs. Different types of LSAs carry different pieces of the puzzle:
- Type 1 (Router LSA): Generated by every router, listing all of its directly connected links and their costs. Flooded within a single area.
- Type 2 (Network LSA): Generated by a designated router on shared network segments, describing which routers are connected to that segment. Also stays within a single area.
- Type 3 (Summary LSA): Created by border routers that sit between areas, carrying condensed routing information from one area into another.
- Type 5 (External LSA): Created by routers that connect the OSPF domain to outside networks, advertising destinations learned from other routing protocols. These flood across the entire OSPF domain.
All of these LSAs are collected into what’s called the Link State Database (LSDB). Every router within the same area must have an identical copy of this database. That synchronized view of the network is the defining characteristic of a link-state protocol.
How OSPF Calculates the Best Path
Once a router has a complete topology map, it runs Dijkstra’s Shortest Path First (SPF) algorithm to compute the cheapest route to every destination. The algorithm starts at the router itself and works outward, evaluating each link’s cost and building a tree of shortest paths. The result is a routing table where every entry represents the lowest-cost path through the network.
The computational cost of this calculation scales with the size of the network. For a domain with 200 routers, early benchmarks on 1990s-era hardware clocked the SPF calculation at around 15 milliseconds. Modern routers handle much larger networks comfortably, though memory remains the practical ceiling. Each external LSA consumes roughly 64 bytes of memory with its supporting data, so a database with 10,000 external entries uses about 640 KB. That was a meaningful constraint decades ago but is trivial on current hardware.
Why Link-State Converges Faster
Convergence is how quickly every router in the network agrees on a new set of best paths after something changes. Link-state protocols like OSPF converge significantly faster than distance-vector protocols because topology changes are flooded directly to all routers, rather than being passed along one hop at a time. When a link fails in OSPF, every router in the area learns about it almost simultaneously and recalculates independently. In a distance-vector protocol, upstream changes take longer to resolve because updates ripple through the network neighbor by neighbor.
OSPF can push convergence even further with fast hello packets. By default, OSPF sends hello messages every 10 seconds on most link types, but it can be tuned to send multiple hellos per second. Setting the dead interval to 1 second and sending 5 hello packets in that window, for example, means a lost neighbor is detected within 1 second. For even faster failure detection, most engineers now pair OSPF with a separate protocol called Bidirectional Forwarding Detection (BFD), which can detect link failures in milliseconds.
Areas: How OSPF Scales the Link-State Model
The link-state approach has an inherent scaling challenge. If every router needs a complete map of the entire network, the database grows with every router and link you add. OSPF solves this by dividing the network into areas. Each area is essentially a self-contained link-state domain. Routers within an area maintain a full topology database for that area only, with no detailed knowledge of other areas’ internal structure.
Border routers that connect areas pass summarized routing information between them using Type 3 LSAs. This means a change deep inside one area doesn’t force routers in other areas to recompute their entire SPF tree. The topology flood stays local, and only the summarized result crosses the area boundary.
OSPF takes this even further with special area types. A stub area blocks external route advertisements entirely, replacing them with a single default route. A totally stubby area goes one step further, blocking both external and inter-area route summaries. Routers in a totally stubby area keep only their own area’s topology plus a default route, which dramatically shrinks the link-state database on those routers.
OSPF Compared to Distance-Vector Protocols
The link-state vs. distance-vector distinction is not just academic. It produces real differences in how networks behave. RIP, a distance-vector protocol, selects routes based solely on hop count, so a path through three fast links and a path through three slow links look identical. OSPF uses configurable link costs, typically based on bandwidth, so it can prefer a single high-speed path over multiple slower hops.
Distance-vector protocols are also more prone to routing loops during convergence because routers don’t have a complete view of the network. They rely on mechanisms like split horizon and hold-down timers to work around this limitation. OSPF’s complete topology database makes routing loops inherently less likely, since every router computes paths from the same authoritative data.
The tradeoff is resource usage. OSPF demands more memory to store the link-state database and more CPU to run the SPF algorithm than a simple distance-vector protocol needs. For small networks with a handful of routers, this overhead provides little practical benefit. For networks with dozens or hundreds of routers, the faster convergence and loop-free routing make OSPF the standard choice. It serves as an Interior Gateway Protocol, meaning it handles routing within a single organization’s network rather than between organizations on the public internet.