Advanced Research Project Lays The Groundwork For All- Optical Network of Future

May 19, 2004

By Nick Wreden, News@Cisco

An advanced generation of optical packet routers may finally see the light if a government-funded project overcomes unknown obstacles and maps unexplored technologies.

Funded by a Defense Advanced Research Projects Agency (DARPA) grant, the four-year project seeks to develop optical packet routers that handle data packets without any optical-to-electrical conversion. These routers will carry more than 100 terabits per second (Tbps), or about 100 times the capacity of today's most advanced router.

"Although this represents leapfrog technology in a world filled with commodity chipsets and me-too products, our team also wants to make sure that we provide a migratory path that incorporates today's infrastructure," says Dan Blumenthal, professor of electrical and computer engineering at the University of California, Santa Barbara. The team, known as LASOR, for Label Switched Optical Router, is made up of researchers from UC Santa Barbara and Stanford University as well as from Cisco Systems and other leading technology firms.

Success has tremendous implications for the future of connectivity. By handling data streams greater than 10,000 feature-length films per second, optical routers can easily enable 3-D videoconferencing as well as telepresence, or the ability to see, sense and even smell a remote event. Bandwidth could become "too cheap to meter," enabling a world where almost every household or business device links to the Internet. With plenty of available bandwidth, disaster recovery planning gets simpler. Even the basics of routing might have to be rewritten, or networks architected differently.

Today, the bulk of Internet traffic essentially travels on separate wavelengths within fiber-optic strands. Data travels via photons at close to the speed of light until they hit routers and switching points. Then the light signals must be turned into electronic signals and then converted back into photons for the next leg of the journey. These optical-to-electronic connections represent major bottlenecks. Additionally, the electronic conversion consumes large amounts of power, increases the router footprint, and inhibits scalability.

But an optical packet router, based on a single stamp-sized chip known as a tunable all-optical wavelength converter, can eliminate these constraints. By enabling photon-to-photon copying without electronics, the highly integrated "photon copier" chip increases router throughput, substantially reduces power requirements, and maintains signal quality. Traffic can be routed to any wavelength on any port, increasing switching speed. Latency - the amount of time data packets wait within a router - is trimmed dramatically. In addition to lowering the cost of equipment, the chip will also help achieve one of the project's ultimate goals - to shrink an advanced router that now occupies a 7-foot equipment rack down to a single line card.

Optical packet routers will transcend the limitations of today's optical switches. These switches mainly use a connection-oriented approach, much like a typical telephone connection, routing data based on pre-determined wavelength assignments. As a result, they lack the flexibility of IP routers, which make routing decisions on a per packet basis and provide the scalability and flexibility behind today's Internet.

To move closer to the dream of an all-optical network based on optical packet routers, numerous challenges must be conquered. One is buffering, used today for routing contention and congestion control. "Only a very limited number of photons can be buffered with today's technology," says Blumenthal. "How do you build a router with limited buffering? How do you create a chip-based solution to optical packet buffering? Can it even be done? Finding out what's possible and what might not be is what makes this project so exciting."

Another key challenge is processing packet labels, used to determine the next network destination as well as provide security and QoS functions. Some key issues include: How can electronics process and forward packets at the ultra-high speeds demanded by an optical packet router? Will label processing and router control continue to be electronic? Will new routing protocols be required?

"Cisco in an integral part of this project," says Blumenthal. "Its expertise in this area is unparalleled, and will help us meet our milestones."

"Cisco is excited to bring its routing and optical technology leadership to bear in this innovative research program. Whether through our own research and development efforts or participating in these kinds of collaborative programs, Cisco is committed to helping develop next-generation network technologies," says Prem Jain, senior vice president, Routing Technology Group, Cisco Systems.

Blumenthal emphasizes that the LASOR project is research into the future, and not product development. "This is a high-risk development with no guarantees. We know we are going to hit walls. Is it even possible to scale routers by this order of magnitude? We will learn by doing, and by experimenting with what works and what doesn't," he says. "In many ways, we're at the same point we were in the 1950s just after the invention of the transistor."

Just as the transistor gave birth to the computer age, optical packet routers will form the foundation of next-generation networks. Information will be delivered literally at the speed of light, with almost limitless bandwidth. By marrying imagination with technology, Cisco and the LASOR team are on a mission to light up the future of networking.

Nick Wreden is a freelance journalist located in Atlanta, GA.