eBGP Question Part 2

Mavrick25 · ‎11-25-2008

Guys,

I just got in an argument with my colleague.

This is the senario:

We have to 2 Cisco GSR routers, that connect to 2 7609 Internet Routers in a FULL MESH topology all running ISIS as metric of 6 as our IGP.

The GSR's then point to another autonomous system running only eBGP.

Now, my colleague is insisting that in the FULL MESH topology we have (listed above)..

To achieve proper "traffic balance", meaning that taffic flows evenly through all links.. load balances...

You need to have the nieghorships on the physical links and not pointing to the loopbacks..

Is this correct..??

I am under the impression that:

If you have ISIS properly configured with equal metrics you will achieve load balancing and traffic will flow evenly across all links... And you the neighborships are done with the loopbacks..

Who is correct... I think I am...

Mav

Giuseppe Larosa · ‎11-25-2008

Hello Mav,

I strongly agree with you:

what makes the difference between BGP and internal protocols is the automatic recursion on the BGP next-hop:

traffic for a BGP (including iBGP) learned prefix follows the path(s) of the BGP next-hop.

You can verify this using the show ip cef commands.

If you have iBGP sessions on loopbacks with next-hop-self on the GSRs towards the C7609s or you keep unchanged the eBGP next-hop you are in good position to load-balance:

the paths are those of the BGP next-hop.

do sh ip route BGP-next-hop

and sh ip cef BGP-next-hop

MPLS is based on this fact: MPLS VPN have as BGP next-hop the remote PE's loopback and so on

Hope to help

Giuseppe

Mavrick25 · ‎11-25-2008

Giuslar,

The GSR towards the C7609 are directly connected with the next-hop-self attribute active.

So essentially..

If the GSR only propagate the 0.0.0.0/32 route and the C7609 receive this routes on 2 different links and because the metric are the same it will load balance..

correct.. Just want to rub it in my colleague face.. :)

MAV