Question: I have an ATM to FR hub and spoke. Im running OSPF.
Why use point-to-multipoint subinterfaces? In what situation would one do that? What are the advantages of doing that over creating point-to-point subinterfaces on the hub and spoke routers with dynamic DLCI mapping (Inverse ARP), instead of the configuration overhead of point-to-multipoint with static mappings?
You can do DLCI to IP maping via Inverse Arp ( which learn dyanamically) or you can do via Static maping ( don't forget to use "broadcast" eitherwise routes will not learn dyanmically ) ,it is depend on you which one you want to use.
Point to Multipoint : You can use P2M when you have hub and spoke location where your branch offices are different place geographically. In this topology Spoke to Spoke traffic has to come via HUB location ( Spoke-HUB-SPOKe) this will increase your lentcy time.
Point-2-Point : you can use where you have more interface between geographical location.
using different point-to-point subifs consumes some resources in terms of IP addresses (ok this shouldn't be an issue if you are using RFC1918 private addresses for your backbone links) and also of resources on routers.
A router platform allocates for each interface and subinterface one position from the IDB (interface database).
The maximum number is platform fixed and IOS release dependent but in order of 1000 or 2000.
So if you have an hub and spoke with hundreds of spokes using a point-to-multipoint interface makes the hub router to save some resources.
Of course OSPF could be not able to scale to a similar scenario using a single point-to-multipoint subif.
Or also the point-to-multipoint is the result of choices that have been made in the past and now moving to multiple point-to-point will require an heavy migration because changes will be required at each remote site.
Traffic will go through the hub in both scenarios because a full mesh will be too expensive.
Nowdays for a new installation I will go with point-to-point links