NSF exec brings supers to science

Stephen Elbert, the new director of the National Science Foundation's Partnership for Advanced Computational Infrastructure (PACI), tries to work technology into his job routine as much as possible. His love of technology became apparent when he asked to set up a videoconference for a telephone interview with a reporter.

Elbert uses videoconferencing technology whenever he can, mostly to check in with colleagues across the country. "When you can do a quick videoconference with someone, you feel more trusting of them," he said. "I have a little hundred-dollar quick cam. I could do it right now."

Unfortunately, the reporter's communications infrastructure consisted of an analog phone line and no camera, so the idea was abandoned. But later, the conversation returned to the subject, and Elbert described video and wireless communications as the most unexpectedly "fun" and most useful new technologies that he has encountered.

E-Mail on the Metro

He mentioned that he now has a wireless modem on his laptop. "If they could just get the signal into the [subway], that would be great," he asserted, savoring the possibility. "So I can hop on the Metro and still be in touch to process my e-mail."

Although he spent most of his career as a computational chemist at Iowa State University's Ames Laboratory in his hometown of Ames, his interests gradually turned from innovative chemistry to creative computing.

Elbert's job requires him to envision new applications for information technology. PACI, which he took over in May, distributes $65 million annually to universities to buy some of the fastest computers and networks in the world, which are used for scientific research.

Physicists, chemists, climatologists and astronomers— among others at the cutting edge of their fields— push these systems to their limits. And the innovations they devise to solve scientific problems eventually trickle down to the average desktop.

"Basically, no amount of computer power is ever going to let you solve the problem you really want to solve," said Elbert, 52, describing an equation used in theoretical chemistry, the field in which he received his doctorate 25 years ago. "There's not a way we can come up with a straightforward solution to it. It's just a matter of how clever you can be in getting computing machinery to operate."

Harnessing Supercomputing Power

PACI is the follow-on to the NSF Supercomputer Centers program, which ended last year. More universities and federal laboratories are part of the new program, which aims to harness the power of supercomputers throughout the nation for collaborative projects.

"We're right on the edge of defining a whole new kind of computing, where supercomputers are a large distributed entity across the country," Elbert said. "I think we can offer the academic community a level of performance that has been unthinkable until lately."

Supercomputers at some PACI-funded institutions were among 13 systems used recently to run a Defense Department battlefield simulation project that utilized 100,000 virtual tanks, airplanes, trucks and other military entities. The institutions were able to complete the simulation four years before DOD expected to be able to run such an application.

Elbert quickly ticked off half a dozen other problems that he thinks more powerful, interconnected computers can solve, such as improving weather forecasts, inventing new drugs and developing more efficient automobile engines.

He said this "partnership" approach to computing is the paradigm of the future. "You'll have very powerful systems on your desk, linked to more powerful systems down the hall, linked across the campus, linked to national or regional facilities," he said. "At some point, you won't need to think about where the work you're trying to do is being performed."

The challenge of forging these partnerships lured Elbert away from his job with the Energy Department, where he had become program manager for Grand Challenge Applications, a project that uses high-performance computers to research problems with broad scientific and economic impact. He also held the position of managing development of DOE's Energy Sciences Network and related projects from 1992 to 1993.

"I find the rapid evolution of technology at all levels really exciting," he said. "I feel like a kid in a toy shop. I want to make sure we continue to make use of this technology in intelligent, meaningful, useful ways."

Mainly, Elbert concluded, "it will be in ways we haven't yet thought of."