Networking’s Big Names Feel the Heat in Race to Move Data Faster

Along the pure glass byways of optical fiber that make up the core of the world’s voice and data networks, a traffic jam of huge proportions is building.

Technology is quickly reaching the point where each hair-like strand of glass--within a bundle of dozens to hundreds of strands--will soon be able to carry a trillion bits of information per second, a blazing burst of data that could represent millions of simultaneous phone calls and Internet data packets.

But the junction points of the network--the brains that route the information to its destination--can currently handle only several billion bits of information per second, an adequate speed for now, but a grim benchmark for the future.

“We’re in a situation now of running as hard as we can to just stay a little ahead,” said Craig Partridge, a chief scientist with BBN Technologies, an information technology research and consulting firm. “The demand for more bandwidth is growing exponentially.”

The perpetual threat of data gridlock, which has become critical during the last few years of the Internet explosion, has launched an intense and contentious race to build what is known as the terabit router.

The biggest names in data and telecommunications networking, including Cisco Systems, Lucent Technologies and Nortel Networks, have started their own projects to build the next-generation machine, which will sit at the center of what is known as the core of the Internet.

But the leading edge, as is often the case with high-end technology, has been seized by a small group of start-ups that are focused exclusively on the terabit prize. Avici Systems Inc. of North Billerica, Mass., Nexabit Networks Inc. of Marlborough, Mass., Pluris Inc. of Cupertino and NetCore Systems Inc. of Wilmington, Mass. are considered among the main contenders in the race, according to some analysts.

Juniper Networks and Argon Networks (now part of German telecommunications equipment maker Siemens) are other young companies that are looking at terabit systems, but have focused more on the current market for advanced super-gigabit routers that can handle billions of bits of information per second.

Even with such a small group of companies, the waters are churning with activity. Telecommunications equipment maker Ericsson of Sweden just announced its purchase last month of Torrent Networking Technologies Inc., a company that produces multi-gigabit routers, for $450 million in cash.

Nortel, in an effort to cover all its bases for the future, has taken a stake in both Juniper Networks and Avici Systems, along with pursuing its own research into creating the next-generation router.

Juniper Networks filed documents last month with the Securities and Exchange Commission for an initial public stock offering.

While there are still years to go before such devices reach the mainstream of the public network, the first real-world tests of these devices this year mark the opening of a potential crack in a market that has been completely dominated by a single company--Cisco Systems.

“The requirement to rebuild the core network is an inflection point and it creates the opportunity for companies to stake out new turf,” said Joe Skorupa, director of switching and routing research for RHK Inc., an information technology research and consulting firm based in South San Francisco. “This is a chance for someone to shake up the apple cart.”

The move toward terabit routers has been driven by a push and pull of forces that have been building over the last half a decade since the advent of the public Internet.

The push side has been driven by the ever-increasing demand for Internet and other data services that has taken off with an exponential trajectory.

The estimated 50 million users of the Internet have pumped the network into an oversized behemoth that just surpassed the telephone system last year in the amount of information it carries.

RHK estimates that the demand for more speed and capacity will increase by five times this year and four times next year. By 2002, the demand will have grown by nearly 200 times.

The pull side of the equation has come from a variety of technological advances that have suddenly allowed for the transmission of huge amounts of data.

In the 1970s, the days of Arpanet--the precursor to the Internet--data was transmitted at what was the then-blazing speed of 56 thousand bits per second. That figure was quickly increased over the copper wires of the telephone networks to millions of bits per second.

But the introduction of optical fiber in the early 1990s began to change the nature of the network. From millions of bits per second, the backbone of the Internet became capable of carrying billions and even trillions of bits per second in the form of light pulses.

What allowed the sudden explosion of speed and capacity on glass was a technology, introduced only in the last several years, known as wave division multiplexing, which in essence allowed many streams of information to be sent over a single glass strand.

By using slightly different frequencies--in essence, different colors--multiple streams of information can be sent without interfering with each other. Many advanced backbones today have about 16 streams, or channels, each of which can carry about 2.5 billion bits of information per second.

However phenomenal that speed may seem, it is only the beginning. The public network will probably see deployment of 10-gigabit fiber systems next year, along with an increase in the number of channels that can be transmitted on each strand.

Nortel Networks, for example, just announced this month the introduction of an optical fiber system that will carry data over 160 channels, at a rate of 10 gigabits per second for a maximum capacity of 1.6 terabits per second.

Handling today’s streams of information requires machines that work at very high speeds, converting optical signals to electrical ones, inspecting each clump or packet of information for its destination and then dispatching it along the proper fiber route.

Cisco Systems is by far the dominant force of the core network, encompassing more than 90% of the market, according to International Data Corp., an information technology research and consulting firm in Framingham, Mass. Cisco has achieved this by providing a wide spectrum of routing products that stretch from the very edge of a corporate network to the deepest core of the Internet. Its top-of-the-line 12,000 series routers essentially define the state of the art, handling a maximum of about 27.5 gigabits of information per second.

But analysts and researchers say there is a limit to relying on ever more powerful machines. The development of new chips can progress at a fantastic rate, but not as fast as the increasing fiber capacity of the network.

The manufacturers can buy some time by simply adding more machines and tying them together, but Partridge of BBN Technologies said that ultimately it takes too much effort for all the machines to coordinate with each other. More time is spent getting organized than routing information, he said.

The ideas for jumping beyond the current generation of super-gigabit routers largely grow out of the world of super-computing. The routers use many specialized processors that are electronically linked in a variety of three-dimensional shapes to shorten and simplify the communications paths between them.

“We’ve just taken the architecture that evolved in super-computing and applied it to routers,” said Peter A. Chadwick, Avici’s director of marketing.

One of the concepts behind the routers from Avici, Pluris, Nexabit and NetCore is that they are “scalable,” that is, they can be expanded by attaching more units, but will still function as a single router.

Most of the terabit designs start out as super-gigabit machines that with the addition of more units can eventually reach terabit levels of routing.

“The old architecture is at the end of its life cycle,” said Mukesh Chatter, founder and chief executive of Nexabit. “The rubber band has been stretched far enough and it doesn’t want to go much further.”

*

For companies such as UUNet Technologies, a leading provider of Internet services, the ability to grow a system is critical.

Kevin Boyne, UUNet’s vice president of global technology operations, said that technology is being outmoded at a rapid pace. “A core router placed six to nine months ago is already being replaced,” he said. “Absolutely one of the keys to success is being able to go forward with this next generation of routers.”

Partridge said the need for terabit routers probably amounts to no more than a half-dozen machines.

The pace of sales will pick up in the next few years, analysts say. Paul Strauss, senior analyst with International Data Corp., estimated that the market will have an annual growth rate of 139% from this year to 2003, when it will amount to a $250-million-a-year business.

Skorupa of RHK is more optimistic, predicting a $1.6-billion industry in terabit routers in four years.

“This will be the year when big, scalable routers roll out, but the big year for terabit routers will be next year and the year after that,” said Skorupa of RHK.

*

Time staff writer Ashley Dunn can be reached at ashley.dunn@latimes.com.