[c-nsp] BGP Multipath load balancing

[gscisco@xs4all.nl Cisco]

Well there is a couple of things that you can do but it depends of the
reason of your current setup.
- Terminate multiple access circuits on one router. You can then have one
BGP peering session and it will load balance nicely. Both egress in and
ingress. You would need to peer to 1 upstream provider then.

-Then again you might want multiple (physical) routers to connect to your
upstream in different pop's for resilliency. That way you can do a traffic
analysis and try to find a split even point (you will of course need to
review that to make sure it will continue to work) and then prefer certain
routes from others in BGP (local pref and such). And specify MED's/AS
prepends for incoming traffic.

- a nicer solution keeping you equipped with multiple physical routers for
redundancy would be to run GLBP between the routers. That would load balance
your egress traffic automatically.

I guess it depends on what you want to establish ... and the reason for the
setup as you have it.

Regards,

Vincent

On 5/23/06, David Coulson <david at davidcoulson.net> wrote:
>
> I've got a couple of routers running iBGP between them, each with a eBGP
> session to an upstream provider. I'm trying to figure out a way to
> distribute the load across the two routers a little better (aside from
> pointing different hosts to each router as their gateway)
>
> I'm assuming that iBGP isn't considered the same as eBGP when
> considering if it is appropriate multipath a route, as in the case
> below. Ideally I'd like this route to be load balanced across the two
> routers, rather than just being sent directly out the local router.
> Currently the route preference is done only by AS_PATH length, at least
> that's what I can figure out. It's routes which have identical AS paths,
> and so forth that I'm more concerned about at this point. I've been
> tweaking local-pref stuff and metrics, however it seems like they're
> more of an all or nothing deal.
>
> core1#sh ip bgp 66.111.55.31
> BGP routing table entry for 66.111.32.0/19, version 638669
> Paths: (4 available, best #3, table Default-IP-Routing-Table)
> Multipath: eBGP iBGP
>    Advertised to update-groups:
>       1          2
>    2828 3356 21840
>      207.166.219.2 (metric 2) from 207.166.219.2 (207.166.219.2)
>        Origin IGP, metric 1000, localpref 100, valid, internal
>    2828 3356 21840, (received-only)
>      207.166.219.2 (metric 2) from 207.166.219.2 (207.166.219.2)
>        Origin IGP, metric 1000, localpref 100, valid, internal
>    174 3549 21840
>      216.28.126.133 from 216.28.126.133 (154.54.4.47)
>        Origin IGP, metric 1000, localpref 100, valid, external, best
>        Community: 174:21000 174:22013
>    174 3549 21840, (received-only)
>      216.28.126.133 from 216.28.126.133 (154.54.4.47)
>        Origin IGP, metric 137980, localpref 100, valid, external
>        Community: 174:21000 174:22013
>
> Any information which anyone can provide me with to get my head around
> this would be great :-)
>
> David
>
>
> _______________________________________________
> 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/
>