From: Matt Bentley (mattdbentley@gmail.com)
Date: Fri Jul 25 2008 - 10:05:38 ART
Hi GS:
I found a unique way to load balance (kind of) without using the bgp dmzlink
command set. If you have a topology as follows:
http://img257.imageshack.us/my.php?image=drawing8ye0.jpg
(Sorry about the popups!)
Basically running OSPF - or any other routing protocol that supports load
balancing between all devices.
BGP Peering:
R8/R9
R8/R6
R4/R6
R6/R9
Created loopbacks 26.26.26.26 on R4 and advertised them into BGP.
On R6/R9, as advertising to R8, I set the next-hop on all prefixes to be
5.5.5.5 - which is R5's loopback.
The next-hop shows as 5.5.5.5 on R8, which is load balanced between R4/R6.
**Note, you do have to set the ebgp-multihop command on R8 peerings to
R4/R6, otherwise, it won't accept the updates.
*Jul 16 20:32:48.091: ICMP: echo reply sent, src 26.26.26.26, dst
172.16.9.53
*Jul 16 20:32:48.151: ICMP: echo reply sent, src 26.26.26.26, dst
172.16.9.50
*Jul 16 20:32:48.207: ICMP: echo reply sent, src 26.26.26.26, dst
172.16.9.53
*Jul 16 20:32:48.267: ICMP: echo reply sent, src 26.26.26.26, dst
172.16.9.50
*Jul 16 20:32:48.327: ICMP: echo reply sent, src 26.26.26.26, dst
172.16.9.53
Of course, the ethernet segment between R4/R6/R9 would be as congested as
ever, but I thought it was a unique application.
Also, for a question, I have been reading through the DOCCD as my lab date
approaches. I really have just stuck with the configuration guides and
command references for the 12.4t and 12.2 versions. Am I just missing out
on a bunch of stuff I need to know for the lab?
This archive was generated by hypermail 2.1.4 : Mon Aug 04 2008 - 06:11:57 ART
