One Man's Dream to Spin a Faster Web

February 8, 1999

One Man's Dream to Spin a Faster Web

By JOHN MARKOFF

ALO ALTO, Calif. -- As Alan Huang sees it, the key to the Internet's future may lie in a simple card trick.
In a restaurant here recently, Huang, a 50-year-old former Bell Labs data network designer, used a new deck of playing cards to help explain how an esoteric branch of mathematics known as group theory offers a simple solution to the coming explosion of Internet traffic.

Annie Dowie for The New York Times
Alan Huang, a former Bell Labs data network designer, holds his model of a multidimensional array of routers, which he says would help ease Internet bottlenecks.
He separated the cards into three piles, then reassembled them into a single deck, which he riffled with a flourish to show that they were in seemingly random order. After several more times of separating them into three piles and reassembling the deck, he uttered, "Abracadabra." As if by magic, the cards had been reordered into the original sequence and suits of the new deck.
Huang offered the card trick, well known among magicians as the Grand Tour, as a simple demonstration of a type of mathematics called group theory -- in which it is possible with a given set of numbers to change their sequence and then reorder them by repeatedly sorting them by using a specific mathematical function known as a generator. But more to his business point, Huang presents the card trick as a symbol for the new method -- an efficient generator -- he says he has discovered for sorting out the welter of data that often collides at the Internet's bottlenecks.
Whether theory can translate to practice, though, remains a very open question. Despite some impressive credentials in computer-networking circles, so far he has found no backers and has depleted most of his life savings in an effort to keep his technological dream alive.
If his technique does work, there would seem to be great demand for it. Almost everyone agrees that greater Internet efficiency will be crucial as vast new torrents of data pour onto the network through new voice and video communications services. What is still subject to debate is how best to accommodate this traffic. Industry leaders like Cisco Systems Inc., the biggest maker of so-called router hardware for directing network traffic, insist that the only way to increase the current Internet's carrying capacity is to build bigger, more expensive routers. But Huang, card deck and group-theory mathematics in hand, contends that he has conceived a novel way to string together the kind of less-expensive routers already widely used on the Internet to expand its data-carrying capacity.
And yet, if the prevention of Internet traffic jams is as deft as Huang's card trick suggests, why have no big companies or rich investors been willing to give him table stakes?
Deaf ears are hardly the reception that one might expect for a researcher with Huang's credentials. As an AT&T Bell Laboratories researcher in 1982, he was the co-inventor of a traffic-routing approach to computer networks called Batcher Banyan switching, one of the first prototype data networks intended to carry both voice and video. For several years it held the world speed record for a data network of its type.

Diagram
Traffic Control
As the Internet has grown, the specialized computers needed to link it together -- called routers -- have grown more powerful and complex. One researcher, Alan Huang, contends that more powerful routers can be assembled from arrays of routers that are less complex and less expensive.

Then, in 1990, while still at Bell Labs, Huang grabbed national attention when he led a research team that built the world's first optical processor -- an information traffic controller for data conveyed as laser light. Though the invention has yet to make its way into a commercial product, many researchers still think that optical switching will become an essential component for the data networks of the future.
But as Huang now seems to be finding the hard way, theoretical research does not necessarily translate into the practical ideas the investors and manufacturers are willing to bankroll.
"We looked at his idea and it was solid research," said Mike Volpi, vice president of business development for Cisco Systems. "But what is lacking is a practical implementation." In other words: What may work in the ivory tower of academic research may not prove viable in the chaotic real world of the Internet.

For Huang, the Bell Labs ivory tower was shaken by AT&T's decision in 1995 to break into three companies and to split the Bell Labs team between two of them -- the remaining AT&T communications company and the new Lucent Technologies equipment company. At the same time, Huang was hearing the Internet's entrepreneurial siren song. And so in 1995 he headed back to Silicon Valley, where he had been an electrical engineering graduate student at Stanford University years earlier.
Huang set up a company, Terabit Routers, that initially pursued the idea of building a super-router based on optical switching. But after about a year, while reading the literature of group theory, a branch of mathematics credited to Evariste Galois, the early 19th-century French mathematician, Huang had an epiphany.
Maybe bigger and more complex was not the answer to Internet routing. Maybe there was a way to draw upon the mathematics of group theory and create the order from seeming chaos that might let data networks operate much more efficiently -- so much so that clusters of cheap, off-the-shelf routers could be harnessed into acting as a new, faster, more powerful type of router.
Since working his router ideas out on paper three years ago, Huang has diligently solicited Silicon Valley's venture capitalists, hoping to convince someone that his idea deserves financing.
But he is finding that esoteric and high-risk technology concepts can be more difficult to sell to venture capitalists at a time when simple, get-rich-quick ideas for Web commerce seem much more in vogue.
"If you have to have mastered advanced physics, it's harder to understand some technologies than the idea of selling wedding stuff on line," said John Shoch, a venture capitalist at Asset Management Corp. in Palo Alto, Calif.
But Web wedding sites are not Huang's only obstacle. He contends that his ideas also run counter to Silicon Valley's infatuation with "big iron" -- ever more complex and powerful routers for coping with the coming deluge of Internet traffic.
"The good news is that this will accelerate the growth of the Net dramatically," he said of his approach. "But this is just about math, and Silicon Valley is used to building faster chips. It's not used to faster math."
Despite Haung's so far disappointing search, his prospects have begun to improve recently. Late last year he received a patent covering his concept for hooking together computer routers, which he refers to as a Galois network, in a tip of the hat to the French pioneer of group theory.
And lately, Huang has found interest and even enthusiasm in some quarters of the industry. He said that British Telecommunications was looking closely at his concept and was planning an experiment to test his theory.
"We've sat down with Alan in some fierce seminars," said Peter Cochrane, head of research for British Telecom. "The common sense underlying his ideas kind of scream out at you."
Still, Huang faces a daunting task in attempting to catch up with the giant Cisco, which today makes the world's highest-performance router, the GSR 12000. Moreover, there are already three start-up companies -- Avici Systems, Nexabit Networks Inc. and Pluris Inc. -- all shooting for the market for the next wave of networking.
Yet, being late isn't always a disadvantage, said Farrokh Billimoria, a research analyst at Hambrecht & Quist in San Francisco who was co-author of a recent report on the future of the routing business.
"If you have a unique idea and you start six months behind, sometimes it's an advantage," he said. "You can start behind and catch up."
Which is precisely what Huang, shuffling his deck, is betting on.

Help/Feedback | Classifieds | Services | New York Today

	Annie Dowie for The New York Times
	Alan Huang, a former Bell Labs data network designer, holds his model of a multidimensional array of routers, which he says would help ease Internet bottlenecks.