Optical Networks Hold the Promise of Incredible Speed and Efficiency

Researchers at Davis work on next-generation technology

By FLORENCE OLSEN

If a conventional T1 network linked the University of California campus here to its medical center in Sacramento, a 15-minute, three-dimensional CAT scan would take 55

ALSO SEE: Colleges Are Watching Carefully as Optical-Network Devices Hit the Market

years to send. With a fiber-optic network and other light-carrying devices, the same transfer would take about five minutes.

Optical-network research drew S. J. Ben Yoo, an associate professor of electrical and computer engineering, to the Davis campus two years ago from a job as a senior scientist at Bell Communications Research. Mr. Yoo is an expert on wavelength converters and optical-label switching, technologies that may someday play a part in research networks and the Internet.

His computer-engineering graduate students work in a 10,000-square-foot "clean room" with a multimillion-dollar photomask "stepper," a machine used in the fabrication of semiconductors for research on all-optical networks. In a laboratory upstairs is a diagnostic instrument so precise that students can use it to detect a single error in hundreds of quadrillions of bits of data streaming by at speeds of 12 billion bits per second and faster.

Experts say all-optical networks could vastly expand the speed and capacity of systems that carry data, voice, and video traffic. The new technology would supplant the hybrid-network scheme on which telecommunications companies currently rely. Today's high-speed telecommunications networks are not purely optical, and may be inadequate to support both research needs and advances in distance education in the years ahead. The networks use comparatively slow electronic equipment to switch the ultrafast laser-light signals carried over fiber-optic cables.

At Davis, Mr. Yoo leads a team of more than a dozen graduate and postdoctoral students doing research to develop an all-optical router for telecommunications networks. The router would redirect optical "packets" of information to their destinations, just as today's Internet routers relay electronic packets of data. But the optical router would do it so much faster -- at the speed of light. An all-optical router, in theory, would be more efficient in a fiber-optic network than an electronic router, which has to convert light signals into electronic signals before it can read labels and route the packets. That conversion takes extra time, limiting the speed of fiber-optic networks.

Other universities and large telecommunications companies are interested in the research program at Davis -- and a handful of other research institutions -- because it could offer telecommunications networks far greater capacity than they have now. The networks would also be less complicated to operate and less expensive to set up. Some optical-switching equipment is already available commercially, but is slower than the devices Mr. Yoo and his students are working on.

Internet traffic, like that on campus-computer networks, has been doubling in volume every year. The growth rate is outstripping the processing power of computers and routers, which is doubling only every 18 months, says Andrew Odlyzko, head of mathematics and cryptography research at AT&T Laboratories. At such a pace, the amount of Internet traffic carried on global telecommunications networks is expected to exceed the volume of voice traffic by 2002.

Some experts now fear that electronic equipment will reach a limit beyond which it cannot handle the anticipated volume of Internet traffic. The solution, they believe, is in photonics, the technology for generating and controlling light.

Most of the national and regional backbone telecommunications networks already comprise optical cable, and many use a technique called wavelength-division multiplexing, which vastly increases the amount of information that can be sent over the cable's fibers. It simultaneously sends thousands of information-carrying optical signals along different wavelengths in a single fiber, then separates the signals at their respective destinations.

But bottlenecks can develop at any routing-and-switching point, where light signals must be converted back into electronic signals. When the routers become congested, Internet traffic slows down.

Current electronic routers are unable to handle information transmitted at speeds greater than one trillion bits per second, which is far less than the capacity that network engineers expect will be required in the future for public Internet traffic, in addition to new academic-research applications.

For example, in their search for an elementary particle called the Higgs boson, some university researchers say they need to move massive amounts of high-energy-physics data between their universities at the unheard-of speed of quadrillions of bits per second.

At Davis, Mr. Yoo's interdisciplinary research on optical communications is aimed at creating generations of optical routers capable of such speeds. His router technology can read optical labels -- which give the packets' destination -- and relay up to 42 quadrillion bits of lightwave data per second. "It's very exciting," says Mr. Yoo.

"Our optical router can handle many different signals from many different fibers," he says. What's more, the new routers will use much less electricity than current versions do. "There's no point in wasting power," he says.

Mr. Yoo's router uses wavelength converters to avoid congestion in a way analogous to switching lanes on a highway. If one wavelength is crowded, the device simply converts a passing optical stream to another wavelength. "Without having to wait, you switch lanes and go," he says.

He uses wavelength converters fabricated from high-speed semiconductor material that can be made in large quantities fairly inexpensively. Each strand of optical fiber can carry thousands of wavelengths, and each wavelength, he says, has a capacity greater than that of Abilene, the network operated by the Internet2 consortium.

His next, or fourth-generation, router, which Mr. Yoo says could be ready five years from now, will have switching components that are 10 times faster but use less power than the technology currently in his lab. The router table, which lists the available network routes and their conditions, "will be much smaller and more efficient," Mr. Yoo says.

Mr. Yoo is one of 107 faculty members, on four University of California campuses, who are involved in an effort to build an experimental high-speed optical Internet that will connect Davis to the campuses at Berkeley, Santa Cruz, and Merced. The research network will support a new Interdisciplinary Center for Information Technology Research in the Interest of Society, located at Berkeley and financed by $100-million from the state and pledges of more than $170-million from corporate and private donors.

The focus of the center's research will be large-scale information systems to support distance learning; "smart" highways for optimizing the flow of cars and trucks; emergency preparedness; and medical-alert and environmental monitoring. A number of the projects will involve collaborative "many-to-many" applications, which demand unprecedented amounts of network capacity, Mr. Yoo says. "We're looking at many classrooms of many students working with many faculty members at many different places."

The experimental network will use Mr. Yoo's optical routers. Because they can handle packets of any size, he says, they will perform better than current electronic-network equipment at routing data, voice, and video information across shared regional networks. Mr. Yoo's technology gives priority treatment to extraordinarily long "packets," like three-dimensional medical images sent in real time.

Recent breakthroughs in optical-network research are substantial, judging from the papers at a recent Optical Fiber Communication Conference, in Anaheim, Calif., attended by more than 40,000 researchers and others from around the world. University of Berlin scientists described a digital polymer switch for relaying laser-light signals in a telecommunications network. Researchers from the University of Bath, in Britain, reported on their work developing a holey fiber, which transmits laser light through a hollow core rather than solid glass. Hundreds of companies, many of them with academic roots in universities, have sprung up in the past few years to capitalize on optical-network research.

Optical-network technology is a difficult undertaking, particularly compared with electronics, says Robert W. Lucky, corporate vice president for applied research at Telcordia Technologies Inc., of Morristown, N.J.. Light behaves quite differently from electricity, he explains. "Electrons are charged particles, and they can interact with things. It's hard to do processing with light, because the photon doesn't want to interact with other stuff."

"I'm sure these problems will all be solved eventually," he adds, "but they are really hard physics problems."

He expects to see all-optical network devices on the market in three years. "It's like the electronics industry back in the 1970's: It's a tremendous opportunity, because this is the ground floor."

Easy-to-print version

E-mail this article