Re: BGP Confederations

[Robert Raszuk]

Hi ccie10,

Have you considered reflectors ? The configuration is trival and
migration easy. Is your reason for confederations partitioning based on
the need to run multiple IGP instances in your network ? If so what IGP
are you running today (having full mesh of ibgp) and how big is the
network we are talking about here ?

I am asking as I wonder why scaling ibgp mesh with reflectors was not
chosen as the prefered option.

R.
 
> It's not a fun process. I've given the matter quite a lot of thought.
> There are no easy ways to do this. Of course, with sufficient planning,
> some scripts, a good lab environment, and a few hours of downtime, a
> network can be transitioned. I don't recommend it for faint of heart. The
> best candidates for a BGP Confederation design are isolated eBGP networks
> (i.e. networks where AS border routers run eBGP, with no iBGP mesh) and
> greenfield networks. The last network I worked with that made a successful
> transition was Mindspring, AS4355, and they fell into the former category.
> 
> The big question is, how big is the network. It's a trivial transition
> with 10 routers, lets say. If you have 100 fully meshed iBGP peers, I
> wouldn't bet on it. Also, implement communities  - they make certain
> aspects of confederation configuration much easier.
> 
> For more information on confeds, check on the NANOG web site. There are
> several archived presentations on the subject.
> 
> - Dan Golding
> 
> On Wed, 4 Oct 2000, ccie10 wrote:
> 
> >
> > Is there an easy way to migrate an existing network running IBGP full mesh
> > to a
> > Condederation based configration. Any documentation that I have come across
> > states that current BGP configs need to be redone and all peering
> > relationships must torn down and rebuilt. Any addtional info/links regarding
> > confederations would be appreciated
> >
> > thanks