[c-nsp] Load Balancing T1s on 2611

[Rodney Dunn]

CEF doesn't suppport double recursion until after the
CEF changes in 12.2(25)S.

In pre CEF changes it would support one level.

ie: A points to a next hop of B and B points to
    a directly connected ip address.

Rodney

On Sun, Jul 24, 2005 at 04:16:37PM +0200, Oliver Boehmer oboehmer" wrote:
> Pete Templin <> wrote on Sunday, July 24, 2005 2:22 PM:
> 
> > Roger Weeks wrote:
> > [snip]
> >> Here's what I've done:
> >> 
> >> ip cef
> >> ip load-sharing per-packet (on the serial interfaces)
> >> 
> >> dual ip route 0.0.0.0 0.0.0.0 upstream router entries
> >> 
> >> This is all we've done before to get load sharing working.  However,
> >> looking at the bandwidth utilization on the T1s, the old T1 is taking
> >> all the load except for about 20-40k on average on the new T1.
> > 
> > Are the ip route entries pointing to IP addresses or interfaces?  I've
> > found that if they point to addresses, load-sharing does not occur
> > (which can be verified with "sh ip route 0.0.0.0" where only one route
> > is visible).
> 
> But CEF will still recurse to both links if this ip address is resolved
> via both links, i.e.
> 
> ip route 10.0.0.0 255.255.255.0 2.2.2.2
> ip route 2.2.2.2 255.255.255.255 Serial0
> ip route 2.2.2.2 255.255.255.255 Serial1
> 
> will still use both links (I think CEF recurses several steps down, so
> this will work as well:
> 
> ip route 10.0.0.0 255.255.255.0 2.2.2.2
> ip route 2.2.2.2 255.255.255.255 3.3.3.3
> ip route 3.3.3.3 255.255.255.255 4.4.4.4
> ip route 4.4.4.4 255.255.255.255 Serial0
> ip route 4.4.4.4 255.255.255.255 Serial1
> 
> 	oli
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/