PS I keep wondering why people keep reinventing virtual circuits.
A key observation here is that the point of an optical cross connect is to provide a real circuit, not a virtual one. An optical cross connect, functioning along with IP routing and an intelligent traffic management system can be used to dynamically place bandwidth where it is needed, when it is needed. The optical plane provides an active provisioning fabric, allowing the network to be more efficient. And a more efficient network makes for a more profitable ISP. I don't see optical cross connects as an opposition to IP technology. Rather, it provides one of the key means of automating the network that is sorely needed. A much better question to ask is: can IP routing possibly survive its projected growth curve without the enabling technology that a flexible optical fabic provides? Yours in dissent, Tony p.s. Just in case there's any confusion out there, I'm still the world's biggest proponent of IP routing. I just don't assume that we know everything about networking already. I hope that it can be made better.
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. The current ODSI work has the concept of dynamic provisioning completely upside down, IMHO. Bora Tony Li wrote:
PS I keep wondering why people keep reinventing virtual circuits.
A key observation here is that the point of an optical cross connect is to provide a real circuit, not a virtual one.
An optical cross connect, functioning along with IP routing and an intelligent traffic management system can be used to dynamically place bandwidth where it is needed, when it is needed. The optical plane provides an active provisioning fabric, allowing the network to be more efficient. And a more efficient network makes for a more profitable ISP.
I don't see optical cross connects as an opposition to IP technology. Rather, it provides one of the key means of automating the network that is sorely needed. A much better question to ask is: can IP routing possibly survive its projected growth curve without the enabling technology that a flexible optical fabic provides?
Yours in dissent, Tony
p.s. Just in case there's any confusion out there, I'm still the world's biggest proponent of IP routing. I just don't assume that we know everything about networking already. I hope that it can be made better.
X.25 ATM MPLS OWR (Optical Wavelength Routing - as I've seen it) If you've been a reader of things like IEEE Communications, you'll have seen (over the last few months) some articles regarding dynamic control of wavelength routes and routing, and "appropriate" interfaces into IP for that. I suspect that this will see the same level of success that did ATM SVCs. Even if you do MPLS badly enough, you sort of have IP controlled VCs. I just don't see IP having enough (i.e. any) "downward hooks" to control the media layer well at all. At least MPLS is sort of a co-resident control layer -- but then we have data/routing types working that instead of Telco types. I did DDN X.25 for a little while, and wasn't going to fall for ATM SVCs so quickly ... Bizarrely enough, last fall IEEE Communications *did* have an article comparing/contrasting SONET vs. ATM vs. IP convergence on failure, and issues regarding the interaction of having all layers attempting reestablishment simultaneously (though the author seemed limited to knowing only RIP). I can only assume that the fellow was drummed out of the union for such heresy. Dynamic re-routing and/or recovery of wavelengths, but I do not believe that it needs to interact with the IP layer at all (this does not imply that it must not be aware of it). Not everyone seems to get the idea of layering protocols ... - TP On Tue, 9 May 2000, Bora Akyol wrote:
Date: Tue, 09 May 2000 13:41:41 -0700 From: Bora Akyol <akyol@pluris.com> To: Tony Li <tli@procket.com> Cc: nanog@nanog.org Subject: Re: Optical Crossconnects and IP
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.
The current ODSI work has the concept of dynamic provisioning completely upside down, IMHO.
Bora
Tony Li wrote:
PS I keep wondering why people keep reinventing virtual circuits.
A key observation here is that the point of an optical cross connect is to provide a real circuit, not a virtual one.
An optical cross connect, functioning along with IP routing and an intelligent traffic management system can be used to dynamically place bandwidth where it is needed, when it is needed. The optical plane provides an active provisioning fabric, allowing the network to be more efficient. And a more efficient network makes for a more profitable ISP.
I don't see optical cross connects as an opposition to IP technology. Rather, it provides one of the key means of automating the network that is sorely needed. A much better question to ask is: can IP routing possibly survive its projected growth curve without the enabling technology that a flexible optical fabic provides?
Yours in dissent, Tony
p.s. Just in case there's any confusion out there, I'm still the world's biggest proponent of IP routing. I just don't assume that we know everything about networking already. I hope that it can be made better.
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 ----------------------------------------------------------------------
| 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. The key words in this sentence are "on demand". I believe that we've demonstrated that traffic engineering is a viable and beneficial capability in large scale IP networks. I would agree that anyone attempting to perform traffic engineering with a very small time constant would be pushing the technology past what is beneficial today. Tony
participants (4)
-
Bora Akyol -
Thomas P. Brisco -
Tony Li -
Wayne Bouchard