Businessweek Archives

Math Genius With Lab. Will Work For Food

Science & Technology


When a key German customer of Sun Microsystems Inc. noticed a bug deep in its software in January, Sun needed a fix fast. For this emergency, it drew on a new reservoir of skill: Russia. A few experts in St. Petersburg quickly spotted the errant code, and within a few hours the Germans had their solution. "It worked great," a Sun manager said later in an electronic-mail memo.

That rapid troubleshooting was "one of a number of early payoffs" from research agreements made last year with Moscow's Institute of Precision Mechanics & Computer Technology, says David R. Ditzel, acting director of Sun Microsystems Laboratories Inc. Sun is so satisfied with its experience so far that it formed a second Russian partnership in March--with ELVIS+, a company headed by computer designer (and Elvis Presley fan) Alexander Galitsky.

The breakup of the Soviet Union and Russia's financial collapse have left some of the world's top scientists underemployed, poorly equipped, and willing to work for nearly nothing. A few U.S. companies, including American Telephone & Telegraph, Corning, and United Technologies (table), have seized on this opportunity and funded research facilities in Russia.

A year or so later, they've made some common discoveries. Running a lab 6,000 miles from home is tough. The obstacles range from language barriers to technology-export rules to, in some cases, a communist-bred lack of initiative. Still, U.S. companies are finding that smart tie-ups can amplify the power of their research and development dollars and give them new perspectives on problems. "Success for Sun Labs doesn't necessarily mean a thing that came 100% out of the Moscow lab," says Ditzel. "It could be just a piece of something."

"CHIP-DESIGN MINDS." For Sun, as for the other U.S. labs in Russia, it's early to expect major breakthroughs. After all, lucrative research can't be turned on like a faucet. But experts say Sun is in the right place at the right time. "It is the first computer company to tap into a resource that has been unavailable to the West," says Christopher G. Willard, a European manager at market researcher International Data Corp. "It could be a tremendous boost for them, certainly worth the risks."

The risks, in fact, are small. At minimal expense, Sun has about 100 highly skilled scientists working in Moscow, St. Petersburg, and Novosibirsk. Roughly half are working on short-term projects such as streamlining Sun's compiler software, which converts programmers' code into language that machines understand. The other half are devoted to futuristic projects. One idea: having the lab's math geniuses write compiler software that would someday let computers run 10 times faster, rather than the 10% increment Sun is seeking in the short term. That's not just a dream. Robert W. Klough, managing director of Microsoft Corp.'s office in Russia, calls the Sun team "one of the best groups of software- and chip-design minds in Russia."

The story of Sun's Russian lab begins in 1991, before the failed August coup that propelled Boris Yeltsin into power and dissolved the Soviet Union. Boris A. Babayan, the father of Soviet supercomputing, found himself short of funds and equipment. He had built the Elbrus-2 machine, the supercomputer used in Russian space and nuclear research. But his work on the even faster Elbrus-3 was being stymied by lack of resources. Since he was intent on designing a superfast microprocessor chip that would use some of the architectural innovations he was putting into the Elbrus-3, he approached Germany's Siemens, Korea's Lucky-Goldstar, and America's Hewlett-Packard. All three said no for various reasons.

Sun, by contrast, was receptive. That June, Ditzel spent two weeks poking around Babayan's rundown lab. "If anything," he concluded, "they had a better grasp of the intellectual part of the problem of chip architectures than their Western counterparts do." The crumbling of the "evil empire" helped the talks along by lessening Western officials' worries about technology-sharing. So in March, 1992, Sun hired Babayan and formed the Moscow Center for Sparc Technology, a Sun-funded unit within the sprawling precision mechanics institute on Leninsky Prospekt. The price: somewhere between $260 and $320 a month per researcher. That's absurdly low by Western standards--but 25 to 30 times the salaries paid in rubles to the rest of the institute's researchers.

"METRIC INCH." Once it had a deal, Sun had the wisdom to put Russians on jobs they're good at--ones requiring skill in the "blackboard sciences" of mathematics and computing theory. Deprived of decent hardware (Westerners joke that ill-fitting Soviet chips are made "to the metric inch"), Russians compensated with ingenious software solutions, such as their work for Sun on compilers.

On the other hand, Sun realized that the Russians wouldn't be any help in making software easy to use. Other Western companies have discovered that shortcoming, too. "Researchers here don't yet have an idea of how an end-user works," says Jane Kitson, president of Lotus Development Russia. For now, Lotus hires Russian programmers only to customize its spreadsheet software for the Russian market.

The Russians working for Sun have more latitude--and use it. Babayan says his team is designing a version of Sun's Sparc microprocessor that incorporates many of the innovations he's building into his Elbrus-3 supercomputer. Among other things, he says it would process many instructions simultaneously to achieve blinding speed. Babayan says he's already working bugs out of the design. Ditzel refuses to confirm that such a project is under way. "We're not asking Babayan to design a chip," says Ditzel, "but to help us understand the essence of his approach."

The miracle is that so much can be accomplished with Russia's economy in the tank, its politics in upheaval--and, most intrusively, Western export-control officers hovering around. "They're stricter than our old KGB," complains Babayan. For example, each delivery of a Sparcstation 10 workstation to Russia requires approval, because it's classified as a supercomputer by the watchdog Coordinating Committee on Multilateral Export Controls (COCOM). Says Babayan: "One-third of the time we are working. Two-thirds of the time we are waiting."

DAILY HEADACHE. Even communications are tightly monitored. Sun's California staff isn't allowed to respond to Babayan's design drafts without permission from COCOM. That can take two months. The mnly exempt communications are go/no-go replies. In those cases, says Ditzel, "all we can say is: 'No, try to do it again.'" He estimates that such rules cut the Russian team's efficiency by 90%. State Dept. technical adviser Cesare F. Rosati says export restrictions have been "incredibly" liberalized in the past four years and adds that if Sun has complaints, it can bring them up at meetings of a government-industry liaison group.

While COCOM is a daily headache, a long-range worry is the prospect of a bidding war for Russian talent. In anticipation that East-West arrangements will proliferate, Sergei Kapitsa, a leading physicist at the Russian Academy of Science, has formed a group called Thesaurus that is dedicated

to fighting exploitation of Russian scientists by Western companies. Says Kapitsa: "When multimillion-dollar deals may be based on your knowledge, you should be paid."

Other Western companies, including AT&T, have sensed annoyance among some Russian researchers, who feel they're not getting a fair cut. Sun concedes that Babayan and his team aren't slated for any royalties from products they help create. Instead, Ditzel says they'll just receive bonuses for good work, exactly as regular Sun employees do.

Babayan claims his group is content with the salaries. "If somebody offers us more, then we'll feel underpaid," he says. So far, Sun's Ditzel isn't worried much about that. Then again, that could change--probably the day his Russian brain trust produces the kind of results he's hoping for.Deborah Stead in Moscow, with Robert D. Hof in San Francisco

American Apparel's Future

(enter your email)
(enter up to 5 email addresses, separated by commas)

Max 250 characters

blog comments powered by Disqus