Hello Ahmed,
As you have yourself pointed out very correctly, there are only two types of network technologies recognized by the IS-IS: the broadcast-type network, and the point-to-point type of network. IS-IS has no provision for point-to-multipoint network type that would be the most appropriate for hub-and-spoke deployment, and it cannot be run directly on such networks.
As far as I know, the only way to run the IS-IS on partially-meshed topology is to convert it to a point-to-point type by creating a point-to-point subinterface for each and every virtual circuit. Thus, in a hub-and-spoke topology, a spoke router will have a single point-to-point subinterface leading towards the hub router. The hub router will have one distinct point-to-point subinterface for each spoke neighbor. Obviously, this design has its drawbacks but it is the most straightforward way to run IS-IS on similar topologies, and it is also the recommended approach.
Using the OSPF-style approach of staying with the multiaccess-type network while making the hub router to be the DR (or the DIS in the IS-IS terminology) and tweaking the IP/DLCI mappings for the next-hop addresses on spoke routers will not help here. The IS-IS appears to make an additional check: while the IS-IS database may be consistent (all routers report the adjacency to the pseudonode for the partially-meshed network, and the pseudonode reports the adjacency to all member routers, thus declaring that all the routers are interconnected), a spoke router does not hear the Hello PDUs from other spokes. Therefore, even though the topology database insists they are interconnected together, they are in fact not because the Hellos from other spokes are not being received. Therefore, on a particular spoke router, the other spokes will not be considered in the SPF algorithm, ultimately resulting in incomplete routing tables on spoke routers.
Perhaps other friends here know a workaround for this but as for my perspective, the only way to make the IS-IS work on partially-meshed topologies is to configure them in a point-to-point fashion before deploying the IS-IS.
Best regards,
Peter
