Important notes about OSPF Network Types

- OSPF defines six network types:
> Broadcast networks
>> Default network on ethernet and FDDI.
>> Will elect a DR and a BDR.
>> Uses the multicast MAC 224.0.0.5 (0100.5E00.0005) for AllSPFRouters and 224.0.0.6 (0100.5E00.0006) for AllDRouters.
>> There is NO next-hop modification. The next-hop IP remains that of the originating router.
>> Layer3 to layer2 resolution is required.
>> Broadcast networks can’t have unicast neighbors configured.
>> 10 hello / 40 dead-interval.
> Non-Broadcast networks
>> Can connect more than two routers but have no native broadcast capability.
>> Non-Broadcast is the default network type on multipoint frame-relay interface, eg a main interface.
>> OSPF routers on NBMA networks elect a DR and BDR, but all OSPF packets are unicast between each manually
specified neighbor with the "neighbor" command.
>> The next-hop IP is not changed and remains the IP address of the originating router.
>> The default priority is 1, and should be disabled (=0) on ALL SPOKES, to prevent a spoke from becoming a blackhole DR/BDR.
>> 30 hello / 120 dead-interval.
> Point-to-point networks
>> Default on T1, DS-3, or SONET links, point-to-point sub-interface on frame-relay and ATM networks.
>> Uses the multicast destination to AllSPFRouters (224.0.0.5), except for retransmitted LSAs which are unicast.
>> NO DR/BDR election, OSPF configured as per normal.
>> The next-hop IP is that of the advertising router.
>> OSPF ignores subnet mask mismatch on point-to-point links.
>> 10 hello / 40 dead-interval.
> Point-to-multipoint networks
>> Cisco proprietary, and not a default option, but best choice for NBMA networks.
>> Are special configurations of NBMA networks in which the networks are treated as a collection of point-to-point links.
>> Does not elect a DR and BDR, and the OSPF packets are multicast (224.0.0.5) to each known neighbor.
>> The next-hop IP is that of the advertising neighbor.
>> Layer3 to layer2 resolution is ONLY needed for the directly connected neighbors.
>> Non-direct neighbors use recursive layer3 IP routing to reach each other.
>> In addition the endpoints of point-to-multipoint networks are advertised as host routes instead
of the actual networks. ie /32 in the routing table.
>> 30 hello / 120 dead-interval.
> Point-to-multipoint non-broadcast networks
>> Cisco proprietary, the same as point-to-multipoint, but configured with the additional 'non-broadcast' keyword.
>> No DR/BDR election, uses unicast appose to multicast, to each manually specified neighbor.
>> As a result the directly connected neighbor must be manually defined with the 'neighbor' command.
This command is only required on the one side.
>> The next-hop IP is that of the advertising neighbor.
>> IP routing will be used to establish reachability between devices that are non-adjacent at layer2.
>> Was created to allow for the assignment of

> Virtual links
>> Are used to link an area to the backbone through a non-backbone area. (Also known as a transit area)
>> Can also be used to connect two parts of a partitioned backbone through a non-backbone area.
>> Must be configured between two ABRs of which one must be connected to area 0.
>> The transit area cannot be a stub area, and must have full routing information.
>> The virtual link will transition to the fully functional point-to-point interface state when a route to
the neighbouring ABR is found in the route table.
>> OSPF ignores subnet mask mismatch on point-to-point links.
>> A virtual link is seen as an interface in area 0.
>> All area 0 attributes are inherited by routers attached to the virtual link, including summarization and authentication.
>> To see the cost of using the transit area use "sh ip ospf virtual-link" and refer to 'cost of using'.
>> The cost of the virtual link is the cost of the route to the neighbors interface via the transit area.
> OSPF over GRE
>> OSPF virtual links cannot transit stub areas.
>> If a virtual link over a stub area is required, the only solution is to use a GRE tunnel.
>> The tunnel interface must have a IP address with a network statement in area0.
> Stub/loopback networks
>> Default for loopback interfaces.
>> Assumes only a single attached router. OSPF advertises stub networks as host routes(/32).
>> Don’t confuse this with stub areas!