I have/am-building this ring of switches at 4-5 different sites, connected by various forms of 10G and using OSPF as an IGP. The basic notion is that each site is its own OSPF area, summarizing the netblock for that range at the ABR. The ring itself is area 0.
The difficulty I have is that the devices driving the ring (cat6500s) are also in some cases the ABRs - overall it's a very collapsed design. (Think of it as a interconnected mesh of pairs of cat6500s linked at 10G, with some sites having other switches and some sites with all L3 being fanned out off of the 6500s.) This is in and of itself also not a problem - but in order for OSPF to work correctly, I need to either: a) define two PTP links between the 6500s - one for area 0 and one for the site area; or b) define the PTP between the 6500s as being in the site area, and create a virtual link for area 0. Every time I read something about virtual links, it comes with this caveat about how this should only be used as a temporary expedient and it represents a horrible design etc etc etc. OK, I accept that it's horrible, but having two separate-purposed 10G links between the 6500s (phys or logical) is ugly too. I can't find a single document anywhere describing this kind of scenario - in fact, it seems as though all of the "case studies" and other example setups all assume a single ABR. Boy, I wish I could have that much faith! Is there some better way to handle this? Or do I just do the virtual-links/dual-connects and accept the hack? Thanks, -bacon _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
