I agree that a more dynamic optical infrastructure allows an IP network to be established faster and better (in terms of flexibility), but I disagree with the point of view that expects routers to dynamically establish, modify and tear-down circuits to other routers on demand. First of all, the current (IGP) routing protocols don't have a clue on who they want to talk to, they talk to whoever is out there and answers their HELLOs. Secondly, we tried this before (ATM) and it did not work.
Whats being spoken of here is a bit different than ATM. Put simply: Build a network of MPLS tunnels (gah, I can't believe I'm about to say MPLS is usefull..) from router A to router B. When this tunnel goes out to reserve bandwidth, it does so. Litterally. At the optical level. Each tunnel becomes a one way circuit in the traditional sense. The router sends out the request to the muxes and they build cross connects to allow the traffic to flow. And in the case of a failure, SONET doesn't get involved. The tunnel reroutes itself by requesting alternate bandwidth. Of course, its a whole heck of a lot more complicated than that, but thats one of the uses that was touched on at the last NANOG. ---------------------------------------------------------------------- Wayne Bouchard [Immagine Your ] web@typo.org [Company Name Here] Network Engineer http://www.typo.org/~web/resume.html ----------------------------------------------------------------------