[c-nsp] Re: Load-sharing to 18 subinterfaces

[Oliver Boehmer (oboehmer)]

Kim,

in your code, CEF does not do multi-level recursion, so resolving three
parallel BGP paths from three next-hops resolving over multiple
interfaces/adjacencies will not produce the desired results.
If you upgrade to 12.2(25)S, you will get the CEF rewrite which provides
this feature, but it also increases the maximum-paths to 32 (a single
prefix will only use 16 paths, but the next prefix will use the other
16, and so on).

Gernally, you are well aware of the caveats of per-packet load-sharing,
so you might want to investigate other access methods providing the
required bandwidth over a fewer number of links.

	oli 

Kim Onnel <> wrote on Sunday, February 19, 2006 1:07 PM:

> Someone said:
> 
> IF you cant do L2 (ATM IMA nor MLPPP) then its L3, run BGP with
> customer over three diff. loopbacks over your PE ? how does that
> sound ? 
> 
> On 2/19/06, Kim Onnel <karim.adel at gmail.com> wrote:
>> 
>> Its a Cisco Internetwork Operating System Software
>> IOS (tm) RPM Software (RPM-JS-M), Version 12.2(14.4)T,  MAINTENANCE
>> INTERIM SOFTWARE 
>> 
>> We couldnt route more than 8 subnet, with a newer IOS, i was told we
>> can put 10, but we need 14. 
>> 
>> One idea is to divide them each in a different VRF and from there,
>> its no more load balancing, its load sharing, does that make sense ?
>> 
>> On 2/19/06, Kim Onnel <karim.adel at gmail.com > wrote:
>>> 
>>> Physical setup:
>>> 
>>> 18 Customers links ---- DSLAM ---- PE ---- Core
>>> 
>>> On the PE router, we configured 18 subinterfaces for a customer,
>>> load-sharing per-packet under the interfaces isnt working, because
>>> there is a limitation on routing subnets on a number of interfaces,
>>> the simplest solution is to put a diff. router for the customer,
>>> but what other solutions are there other than multilink (CPU
>>> Killer) ? 
>>> 
>> 
>> 
> _______________________________________________
> 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/