The worlds of IP routing and optical switching are fast converging, if last month’s NetWorld+Interop 2000 is any indication.
Optical networks were on the lips and minds of many attendees, from the largest makers of equipment for service providers to the smallest start-ups tucked away in the farthest reaches of Atlanta’s vast Georgia World Congress Center exhibition hall.
For users, the convergence of the IP and optical worlds could affect the speed and price at which next-generation services are offered. Virtually all multimedia IP services being provisioned or planned rely on the integration of IP and optical networks, observers say.
But there are many open issues regarding the convergence of IP and optical. Among them are who should “own” the intelligence of the optical network – router vendors or optical switch vendors? Another is coordination of signaling and control plane functions between routers and optical cross connects.
There are also contrasting viewpoints on how many electrical-to-optical conversions are necessary in an IP optical network, and where in the network these conversions should occur.
“The big question is, what’s going to make the end-to-end decisions, the router or the switch? There are fundamental religious wars here,” says Shaym Jha, a vice-president at Corvis Corp., a maker of optical add/drop multiplexers, switches and other gear.
Corvis believes in keeping optical switching and routing decisions separate, yet making both control plane functions interoperable.
“Switching should be left to the optical network,” Jha says. “IP companies live in the routed world. What doesn’t make sense is the router telling the switch which route to take.”
Yet the brains for optical switches haven’t been entirely baked yet, Jha admits. Switching and routing control functions for optical network bandwidth provisioning as well as route restoration and protection are still evolving in the Optical Internetworking Forum and the International Telecommunication Union, he says. The Optical Domain Service Interconnect coalition is also working on optical signaling standards.
“Control in the optical network is being worked out as we speak,” Jha says.
The Internet Engineering Task Force’s Multi-protocol Label Switching (MPLS) specification is viewed as one key way to coordinate the signaling and control plane functions between IP routers and optical switches. Indeed, some optical standards bodies are devising a Multi-protocol Lambda Switching (MP Lambda S) protocol to engineer traffic in optical networks the same way MPLS steers traffic in routed nets.
In routed networks, MPLS adds a “label” to an IP packet or flow to steer that traffic through the network. MPLS can be used for explicit routing, fast rerouting, “hard” quality-of-service constraints, and for routing with non-unique addresses, such as in setting up private user groups in VPNs, says Yakov Rekhter, a Cisco Systems Inc. Fellow engineer and one of the architects of MPLS.
Rekhter presented a conference session at Interop on MPLS and MP Lambda S.
Similarly, optical wavelengths – or different colours of light – can be used as the “labels” for steering traffic through optical networks. As a control plane protocol for optical cross connects, MP Lambda S can be used to provision optical channels, facilitate dynamic, reconfigurable networks, and as an integration point for optical cross connects, wavelength division multiplexers and routers, Rekhter says.
Indeed, the same signaling mechanisms that are used in routed MPLS – the Resource Reservation Protocol and Constraint Routing-Label Distribution Protocol – can be used to set up and tear down paths in optical networks, he says.
“Label-switched paths can span both routers and optical cross connects if they share the same control plane,” Rekhter says. “[MP Lambda S] trivializes control coordination problems among network elements [and] simplifies hybrid administration” between routers and optical cross connects, he says.
Corvis’ Jha says it’s premature to usher in MP Lambda S.
“You should have one control plane in the optical network, one in the routed network, and they should interoperate,” he says. “MP Lambda S will happen but perhaps not in its current form.”
Corvis has demonstrated interoperability between its optical gear and routers from Cisco Systems Inc., Juniper Networks Inc. and Avici Systems Inc. The demonstration showed how the Corvis optical network can steer routed traffic – from up to 160 routers on one fiber, Jha claims – to a destination.
Control plan issues aside, conversion issues – optical to electrical, and optical to electrical to optical (OEO) – are also generating discussion in the IP/optical realm. Converting electrical impulses to light, and light to electrons, can introduce performance delays in the network and make equipment and services more expensive to procure and provision.
VIPswitch, a maker of terabit-capable Ethernet switches for metropolitan-area networks (MAN), says the OEO issue is one for long-haul networks.
“In the MAN, it’s a different issue – cost,” says Yves Hupe, director of product marketing for the Montreal switch maker. “What is the cost point between going electrical vs. going optical? Up to 10 terabit/sec, electrical is efficient. Beyond that point, optical has a definite advantage.”
Others are anxious to turn up the light on the metro and access infrastructure.
“Why should we go through all of the intermediate mappings?” asks Paul Zalloua, director of product marketing for LuxN, a maker of optical access platforms designed to extend optical services from the MAN point of presence to enterprise customer premises at 2.5Gbps. “Our goal is to map everything in light. Why not take T-1 into optics?”
Zalloua says another issue to be worked out in the convergence of IP and optics is network monitoring. Some of the same constructs used to monitor traffic in the packet world will not be suitable in the wavelength world, he says.
“Counting packets dropped and received is not going to scale up,” he says.