Why this matters
The OSPF network type of an interface decides three things at once:
whether a DR/BDR is elected, which timers are used, and how
Hello packets are addressed (multicast vs unicast). A mismatch here
is one of the most common reasons an adjacency silently fails to form.
The five network types
| Network type | Hello | Dead | DR/BDR? | Hello sent to | Typical media |
|---|---|---|---|---|---|
| Broadcast | 10 s | 40 s | Yes | 224.0.0.5 / .6 | Ethernet, Token Ring, FDDI |
| Point-to-point | 10 s | 40 s | No | 224.0.0.5 | HDLC, PPP |
| Non-broadcast (NBMA) | 30 s | 120 s | Yes | Unicast | Frame Relay, X.25, ATM |
| Point-to-multipoint | 30 s | 120 s | No | 224.0.0.5 | NBMA / DMVPN configs |
| Point-to-multipoint non-broadcast | 30 s | 120 s | No | Unicast | NBMA variants |
Loopback is a sixth (internal) type
A loopback interface is advertised as a /32 host route and never
forms adjacencies.
Timers
Default relationship
Dead interval = 4 × Hello interval by default. Broadcast and
point-to-point share 10/40; the other three share 30/120.
Timers must match to form an adjacency
Hello and Dead must be identical on both neighbors. Timers are
configured per interface, not globally — two routers in the same
area can fail to peer over a single interface mismatch.
The 4× ratio is only a default
If you change the Hello interval manually, don’t assume Dead
re-adjusts to 4× on its own — verify it, and remember it must still
match the neighbor.
DR/BDR by network type
Who elects a DR
Only broadcast and NBMA elect a DR/BDR — these are the
multi-access types where many routers share one segment. Point-to-
point and both multipoint types do not: there’s no benefit to a
DR when links are effectively one-to-one.
Hello addressing — multicast vs unicast
Why some types use unicast
Broadcast, point-to-point and point-to-multipoint send Hellos to the
multicast 224.0.0.5. The non-broadcast types (NBMA and
point-to-multipoint non-broadcast) can’t rely on multicast on the
underlying media, so neighbors are defined manually and Hellos
are sent unicast.
Point-to-multipoint and DMVPN — context
Where point-to-multipoint shows up in the real world
Point-to-multipoint is commonly used over DMVPN and other NBMA
hub-and-spoke designs, where treating the tunnel as broadcast would
force an awkward DR election (the hub would have to win it). Setting
OSPF to point-to-multipoint avoids the DR entirely and treats each
spoke as a neighbor. DMVPN itself is a CCNP-level WAN topic — here it
just explains why this network type exists.
Compatibility between network types
What actually breaks or allows an adjacency
Two neighbors can form an adjacency only if their interfaces agree on
the three things a network type controls:
- Matching timers (Hello/Dead) — broadcast↔point-to-point can
peer because both are 10/40; broadcast↔NBMA cannot, 10/40 vs 30/120.- Same DR expectation — a type that elects a DR paired with one
that doesn’t leads to inconsistent behavior.- Same discovery method — multicast on one side, manual/unicast
on the other won’t find each other automatically.The network type can be changed per interface (
ip ospf network <type>) precisely to align these when the media default doesn’t fit.