On Thursday, May 26, 2011 09:59:35 PM Gert Doering wrote: > Try building a fully-meshed network of 2500 routers and > be surprised on the amount of computational power you'll > need...
Indeed. You don't only get to worry about iBGP and how much the route reflectors can scale (either as a cluster or as individual units), but you also need to be very concerned about the "speed" of the IGP. The suggestion from Adam is basically scaling large routed networks using BGP Labeled Unicast. The general idea with this solution is to break the network up into several regions, much like you would a multi-area OSPF or multi- level IS-IS network. However, there's a twist... With BGP Labeled Unicast for scaling large routed networks, each region wouldn't exchange routing information via the normal methods you would in the same AS, i.e., IGP, LDP, RSVP, e.t.c. Instead, the regions would exchange routing information via BGP Labeled Unicast. In all fairness, I think this solution, while valid, is rather complex. Very complex. However, when you're talking about a network that large in this modern day, it might not necessarily be such a bad idea to consider this solution. Cheers, Mark.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
