Research team nets $6.3 million grant for optical router research

BY PAUL DESRUISSEAUX AND BILL GRANT

The Defense Advanced Research Projects Agency (DARPA) has awarded a $6.3 million, 4-year grant to a team of researchers in industry and higher education, including Stanford and led by a group at the University of California-Santa Barbara (UCSB), to develop new technologies to advance optical router capacity far beyond the current state of the art.

The team expects to develop and demonstrate all-optical technologies and systems that route data packets -- the currency of the Internet -- with no optical-to-electrical conversion. The potential payoff of avoiding optical-to-electrical conversions is to greatly increase the data speed and significantly reduce power requirements over today's approaches. This groundbreaking collaboration is expected to open new possibilities for the distribution of rich data, voice and video content.

"Imagine a data stream greater than 10,000 feature-length films blasting through an optical router in one second," said Daniel Blumenthal, a professor of electrical and computer engineering at UCSB and leader of the research team. The research, he explained, will seek "to revolutionize optical integration density and develop new technologies to advance optical router capacity beyond 100 terabits per second," or about 100 times the capacity of current state-of-the-art routers.

The team, known as LASOR (for Label Switched Optical Router), is made up of researchers from Stanford and UCSB as well as several leading technology companies -- Agility Communications, Calient Networks, Cisco Systems and JDS Uniphase.

Stanford researchers will help design the overall architecture of the router. They also will investigate the design of networks of routers with little or no buffering.

"Internet routers typically have very large packet buffers to hold packets during times of congestion," explained Nick McKeown, an associate professor of electrical engineering and computer science at Stanford. "A backbone electronic router today typically has enough buffers to hold hundreds of thousands of packets. It is widely believed that you need this much buffering for the Internet to perform well."

The amount of buffering used today is well beyond what's possible with optics. Plus no one knows how to realistically buffer photons. But the technology developed as part of this project should make it possible to build small buffers, perhaps with a few dozen packets, McKeown said.

"The question is: Can you design a high-throughput, functioning network if the routers have very small -- or no -- buffers?" asked McKeown, whose group is world renowned for design of Internet routers and supported in large part by Cisco. "Conventional wisdom says no -- but we plan to challenge this perception."

McKeown is tackling this challenge with Greg Watson, a senior research engineer, and Yashar Ganjali, a doctoral student in electrical engineering. Ashish Goel, assistant professor of management science and engineering, and Tim Roughgarden, who will join the computer science faculty this summer, are both experts in the design of networking routing algorithms, and bring further world-class expertise to the Stanford team. Since the project involves the design of all-optical Internet routers, over the last three years McKeown's group has worked with Stanford optics experts Professors David Miller and Olav Solgaard on the design of routers with optics inside.

One of the key technologies that will be utilized is the tunable all-optical wavelength converter, an integrated device that acts as a "tunable photon copier" (see www.engineering.ucsb.edu/Announce/photon_copier.html) and is used to direct packets through the router using the color of light itself. The LASOR team's ultimate goal is to shrink the size of state-of-the-art routers that occupy a full 7-foot equipment rack today down to a single linecard. To achieve this goal, the LASOR team will push the boundary of how many optical devices can be integrated onto a single chip, ushering optics from the equivalent of electronics of the 1950s to the electronic revolution of the '60s and '70s.

"Given the historical role of the government in creating the foundation of today's Internet, it is fitting that DARPA would support the development of new technologies that could dramatically affect the future of the Internet," said Prem Jain, senior vice president, Routing Technology Group, Cisco Systems. "Cisco is excited to bring its routing and optical technology leadership to bear in this innovative research program."

Paul Desruisseaux is associate vice chancellor for public affairs at the University of California-Santa Barbara (UCSB). Bill Grant is program manager of the College of Engineering at UCSB. Dawn Levy contributed to this report.