[Babel-users] Kernel metrics/route redistribution

[Will Glynn]

On Oct 22, 2010, at 6:49 AM, Juliusz Chroboczek wrote:

>> Traffic from the wired network into the Babel network should ideally
>> follow the shortest Babel path possible. This would require exposing
>> the Babel distances to the wired network, which is why I started by
>> asking about that.
> 
> Ah.  You want the Babel network to be non-convex.  Ouch.
> 
> I've made my Babel network convex by adding a few tunnels over the wired
> infrastructure.
> 
> I undestand what you're trying to do, and I'm not sure how well it will
> work; I'd expect such a mixture of routing protocols to create
> persistent routing loops unless done carefully, and at any rate to be
> a nightmare to maintain.

Good to know we're around the same page :-)

I think a) giving each network disjoint sets of prefixes and b) making sure Babel never sees re-announcements from itself simplifies the problem to the point where it's workable. That doesn't change anything on the Babel side (it sees default routes being brought in at more than one point), and the IGP side isn't functionally any different than multihoming with different tier-1 ISPs (where prefixes will be closer on one peering session than the another).

It works in my brain, at least. From there, I hope to make it work in my lab, followed by making it work in a trial deployment.

> Notwithstanding that, I'll provide you with a patch to do what you want.

I would appreciate that!

Thanks,

--Will Glynn