[c-nsp] 6500 iBGP mesh

[Bruce Pinsky]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Aubrey Wells wrote:
> Hello,
> We currently have two iBGP peered edge routers that are eBGP peered to 
> several upstreams, as well as several private peering arrangements. 
> These two routers are in seperate data centers connected via a Layer-2 
> ethernet circuit. Downstream from the edge routers is a mesh of 
> Sup2/MSFC2 6506s that are aggregate points for many other POPs. All 
> devices are connected via IGP (ospf). Currently, when a customer route 
> gets to an edge router, it looks up the destination in the table and 
> sometimes has to send the packet over the layer2 circuit to the other 
> edge router. Since all traffic at some point goes through a 6500, I 
> would perfer to have the 6500 make the decision of which edge router to 
> go to for the route, saving me the bandwidth on the L2 interconnection 
> between the edge routers, and eliminating a hop for the customer.
> 
> My thoughts on solving the problem is to set up a iBGP mesh between the 
> 6500s and the edge routers so any one route will (almost) never hit both 
> edge routers. Will the Sup2 be able to handle a full view from each edge 
> router to accomplish my goal? Any suggestions to make it work (outside 
> of upgrading to Sup720)?
> 

I know of a customer that has two full views at about 210K+ routes.  That's
about 170MB of BGP usage.  Depending on which image you are running on the
Sup2 and if you have 512MB of memory, you should be pretty good.  If you
have 256MB you are probably pushing the envelope on the amount of free
memory headroom you have.  Less than 256MB, well, "that dog don't hunt."

One option would be to use another device in the network as a non-transit
route reflector to have a single peering connection between your 6500's
with Sup2 and the RR.  Downside to that is that you have a single point of
failure now instead of having two independent BGP sessions to the
edge/border routers.  And of course adding a redundant RR would put you
right back in the same position as you started.   You'd have to decide
which is more important, the single point of failure or the bandwidth
between the edge routers.

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.4 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFE3Q98E1XcgMgrtyYRArX9AJ9O4HxnlOjvLuI+WbF+vGIU8ueTzACgkqQ3
LwbFH7N6o7t/pv9fttsxNNg=
=UIq2
-----END PGP SIGNATURE-----