The MPLS WG Archive

[Alex Zinin <azinin@cisco.com>]

Bala,

Note that with today BGP that border OXC will be the next-hop
for all IP prefixes it announces to the client via the e-BGP
session. This means that the client does not really know
much about the egress point per prefix and hence where it should
establish LSPs across the optical transport network unless we
add a kind of "suggested next-hop" BGP attribute as a part
of optical UNI (I don't think it makes sense to run e-BGP
in the peer model) to inform the client about the egress
points.

I think Curtis' point is that OXCs do not really need to
have BGP routes in their routing table for normal operation
of circuit provisioning protocols.

Alex.

>> Optical switches are not known for the robustness of their BGP
>> implementations if they have BGP implementations at all.  They may
>> also have more important things to do with their CPU cycles given that
>> they do no externalrouting at all.  It would seem to be a relly bad
>> idea to run IBGP to the optical switches.
>>
>> I hope that was more clear.

> Absolutely not.  It's certainly seems possible to have
> a scenario where optical edge switches run E-BGP
> with router clients to provide reachability information. Whether
> the optical network is a separate AS or a sub-AS is another issue.
> What's wrong with this?

> Bala


> --

> Bala Rajagopalan
> Tellium, Inc.
> 2 Crescent Place
> P.O. Box 901
> Oceanport, NJ 07757-0901
> Tel: (732) 923-4237
> Fax: (732) 923-9804
> Email: braja@tellium.com