OSPF and Frame-Relay Hub and Spoke Networks

Brian,

Before we start configuring any routing protocols why do we, in some of the labs, have reachability between the spokes in a hub and spoke network and other times we don't?

Not having reachability between the spokes by either using additional Frame Relay map statements to the hub on the spokes or using point-to-point subinterfaces on the spokes can cause you to run into "next-hop" issues depending on the OSPF network type used. A walk-through should help you better understand the solution and the problem.

First off lets assume we have three routers in a hub and spoke Frame Relay network*: R1, R2 and R3. R1 is the hub with R2 and R3 as the spokes. The OSPF network type used is either broadcast or non-broadcast as technically it doesn't matter since the next-hop behavior is the same for both network types. In this case OSPF assumes that all routers on the same logical IP subnet have direct layer 2 reachability between each other. If these three routers were on an Ethernet segment they would by default be able to communicate directly with each other via layer 2. Since this is the case when the DR (assume R1) sends an update to R3 about a network advertised by R2 the next-hop for the network being advertised would be R2. This is exactly how you would want the traffic to flow, directly between R2 and R3. If R3 has traffic destined for a network advertised by R2 you wouldn’t want R3 to forward it onto the Ethernet segment to R1 (DR) and then R1 turns right back around and forward it to R2. This would double the network utilization and create an additional hop (R1) that is totally unnecessary.

Now back to the Frame Relay hub and spoke situation. If the OSPF network type is broadcast or non-broadcast as previously mentioned and the spokes routers do not have reachability to each other (no Frame Relay mapping or multipoint interfaces) when R2 advertises a network to R1 (DR) and R1 then in turn sends it down to R3, the next hop for the network will be R2. Since R2 and R3 do not have direct layer 2 reachability (i.e. no DLCI between them) hen they need to reach each otherthey will need to forward traffic to the hub (R1) since R1 is the only router that has direct layer 2 reachability to all routers.

by a with networks advertised by the If you are wondering why spokes need reachability with other spokes, configure a frame relay hub & spoke network using an OSPF network type of broadcast or non-broadcast. Now on one of the spokes look at next hop address for a route advertised by another spoke.

Non-broadcast:
Rack12R4#sho ip route ospf | in 2.2.2.0
O 2.2.2.0 [110/65] via 132.12.0.2, 00:00:13, Serial0/0
^^ ^^^^^^^^^^

Change the network type to point-to-multipoint and then look at the same
route.

Multipoint:
Rack12R4#sho ip route ospf | in 2.2.2.0
O 2.2.2.0 [110/129] via 132.12.0.1, 00:00:32, Serial0/0
^^^ ^^^^^^^^^

Notice how the next hop address changes. The next hop address is now the hub and not the other spoke. Also note that the OSPF cost is more accurate with point-to-multipoint in a hub and spoke environment over using broadcast or non-broadcast. Point-to-multipoint is taking into account the cost to reach the hub and then the cost from the hub to the other spoke.

This Frame Relay network could be substituted with an Ethernet network using private VLANs or protected ports. In the case of the private VLAN example R1 would be in the primary VLAN with R2 and R3 in an isolated VLAN.