On 06/23/2011 06:28 PM, David Barak wrote: >> Here's a quick diagram of what the network needs to look >> like: >> BB1---BB2 >> |\ /| >> | \ / | >> \ POP1 / >> \ / >> \ / >> POP2 >> >> The BBs should be learn all routes in the network (L2), and >> the POPs can be whatever level, but they should only learn >> default, ever. >> > > That type of topology will be problematic in this context - if BB1 and BB2 > are in different areas, then POP1 and POP2 would both be single-homed. If > they're in the same area, then POP1 will see POP2's routes. If the POPs are > L1/L2, they'll definitely see each other's routes. Would it work to use ISIS > for just the loopbacks and interfaces, and then carry all of the other routes > in BGP?
Yes, I've been wondering if our needs are even possible with this topology in IS-IS, hence my hail-mary here. Our goal (and current config) is to only announce loops and interfaces between BB and POPs with IS-IS, everything else is in BGP. This does reduce the amount of prefixes learned, but we're trying to get it as low as possible since the POP routers are 3750MEs, which have limited TCAM, and our network is pretty large (90+ nodes and growing), plus we want to leave TCAM available for v6 in the future. > Alternatively, you could add more routers to the topology so that they could > stay dual-homed in a single area, but that can be expensive depending on the > platform. Unfortunately, inserting new routers between the BB and POP routers to act as L1/L2 boundaries is not possible. > David Barak > Need Geek Rock? Try The Franchise: > http://www.listentothefranchise.com _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
