[c-nsp] BGP/IGP archiecture

[Joe Provo]

On Wed, Sep 06, 2006 at 01:59:00PM -0400, Matt Liotta wrote:
> I have two locations that are each running multiple BGP peers. 
> Originally, the second location was just for backup, but over time we 
> started using the routes at the second location to improve our 
> performance. The problem is that currently all traffic flows through the 
> first location to get to the second location. These two edge locations 
> are connected to several 6500s using a full Ethernet mesh that aggregate 
> all of our traffic. What is the best way to allow the 6500s to do 
> best-exit routing?
> 
> Right now I am looking at either doing an IBGP mesh or aggregating into 
> our existing OSPF cloud. Any thoughts would be appreciated.

Yes. you need an ibgp mesh for all your true core devices (hint, the 
things at the external edge don't have to be part of your true core)
and a way for the ibgp to be set up.  The latter is best done via an
igp distributing loopback /32s and edge [whatevers].  You can start 
out nailing all of that up with startics, but within 1 upgrade cycle
you'll realize you should have gone ahad and used your igp.

-- 
             RSUC / GweepNet / Spunk / FnB / Usenix / SAGE