Intelligent, all-optical 'Net years away
WASHINGTON, D.C. -- Significant work remains to be done if the industry is to realize the full benefits of intelligent optical networking for scaling the Internet and making it more reliable, according to industry visionaries.
At the Next Generation Networks (NGN) conference last week, experts offered sometimes sharply diverging views on how to handle the exploding growth of traffic on the Internet and make it more resilient. They also debated the roles of routers and optical switches in supporting those and other business-critical operations.
The consensus is that the optical Internet is still in its infancy and that much progress needs to occur in standards definition, product development and interoperability. Further work needs to be done in bridging optical "islands" to foster an end-to-end, all optical IP infrastructure to replace today's public network.
"I feel we're at the end of the 'first beginning' of a decade-long effort to rebuild the public network infrastructure," says John McQuillan, president of McQuillan Ventures and chairman of the NGN conference. "There are many remaining challenges in making the 'Net better. We still have work to do."
McQuillan's reference to the "first beginning" indicates an expectation that the industry may attempt to pass through several stages in its quest to deconstruct the public network and rebuild it as the optical Internet.
One of those stages is making the 'Net more scalable and defining the role optics plays, if any. Traffic and bandwidth requirements on the 'Net are doubling every two to six months, pundits say, and differences of opinion exist on how to accommodate that growth.
The role of MPLS
Many believe Multi-protocol Label Switching (MPLS), a technique devised to enable electronic routers and switches to scale the 'Net, can be connected to optical infrastructures. But some believe MPLS is not a scalability panacea.
"MPLS is not the one answer to all problems. It has properties that limit scalability," says Judy Estrin, president and CEO of start-up Packet Design. "The replacement of globally known addresses with [MPLS] tags limits scalability."
MPLS will help resolve the scalability problem by letting routers manage changing traffic patterns and set up faster connections, says Desh Deshpande, CEO of Sycamore Networks. That may provide the intelligence necessary to make the optical Internet a more dynamic environment where bandwidth can be set up and torn down on demand.
But scalability is just the situation optics is designed to address, he says.
"The only problem optics has to solve is the scalability problem," Deshpande says. "If you can create these flexible pipes and fill them up with 10G bit/sec [of bandwidth], that's a scalable model."
MPLS is a complicated technology, and complication could be another factor that's limiting the scalability of the Internet, McQuillan says.
"Service providers can't hire enough people," he says. "It's cool to build the latest, biggest and baddest box, but connecting it to everything else is a limiting factor."
In addition to scale, stability of routing iis another concern in growing the 'Net and making it a reliable infrastructure for daily communication, research and commercial use. Routing table sizes and update intervals are of particular concern, says William Norton, co-founder and chief technical liaison of service provider Equinix.
"The propagation of [Border Gateway Protocol] is a cause for great concern," Norton says, referring to the routing protocol used in the Internet. "You have to wait 15 minutes for updates. Can you imagine talking on the phone and waiting 15 minutes for a response?"
But routing is needed for its ability to inspect packets and request or provision bandwidth. If optics are needed for scale, is an intelligent, all-optical 'Net possible?
"We don't have a way optically to tell you what the bits are," says Stephen Alexander, senior vice president and chief technology officer at Ciena. "If you need to know what the bits are, you need electronics."
Alexander does not believe integrating routing and dense wave division multiplexing is the answer.
"We're not big fans of the 'God box,' " a product that is touted to do everything, he says. "There's a reason your car doesn't fly and float."
McQuillan suggests it may be time for the industry to conceive new routing algorithms that are optimized for scale and optics. Others propose that today's core routers will soon be obsolete.
"Packet routers as we know them today will be dead in 18 months," says Lawrence Roberts, chairman and chief technical officer of Caspian Networks. "They need to be vastly more intelligent to solve the scaling and reliability problems."
A jumble of standards
Compounding these issues is the lack of standards and the proposal of standards that appear to compete or overlap. For example, there are four standards proposed for signaling between routers and optical transmission gear: the Optical Internetworking Forum's User-to-Network Interface (UNI); the Optical Domain Service Interconnect's UNI; Multiprotocol Lambda Switching; and the Automatic Switched Optical Network specification.
"Why four standards and not one? We don't have one language, one currency or one government in the world," McQuillan notes. "But this is a very, very pivotal issue for our industry and we're only at the beginning."