[c-nsp] Alternative to AToM Tunnel Select?

[Adam Strawson]

On Fri, 15 Oct 2004, Eric Osborne wrote:

> By "AToM where each path is explicitly configured", do you really mean you 
> want
> to steer different AToM VCs along their own individual paths, without regard 
> to IP shortest cost?

Yes, thats exactly what I mean though the VC traffic must fail rather than 
re-route if part of its path fails.

> Only way you could do this would involve an ugly set of
> static routes.  There's no way to do something other than IP shortest cost
> without using TE, modulo static routes which of course are not the most
> scalable tool in your toolbox.

I feared that.

> Could you do tunnels on the P routers, and just make sure routes to the 
> remote
> AToM PE point down explicitly routed tunnels on the P?  This would be less
> granular, but therefore less work, than doing it at the PE, and presumably 
> your Ps are 12.0S.

Yes, I think I'll play with this a little using the 3750 Metro as a PE. 
For my requirement to avoid re-routing I'd also likely have to maintain 
statics with a higher admin distance to null route the traffic should the 
tunnel fail.  I'm not sure the extra overhead of the static routes is 
worse or better than the frustration of living with a PE running 12.0S and 
trunking VLANs to some cheap switches as we do today ... but I'll explore it.

Thanks for your thoughts!

Adam.