SuperComputing, What is it?
An Interview with Roscoe Giles
Boston Globe (BG):What is supercomputing?
Giles: Computer power is doubling roughly every 18 months, and what was a supercomputer 20 years ago is now a desktop machine. In recent years, the favored term is ``high-performance computing.'' The technology of high-end computers has essentially merged with the technology of the desktop and PC. Earlier, there was a big firebreak in the technologies, when supercomputers were liquid-cooled and typically cost $20 million apiece. At BU, we have a 192-processor parallel machine [a Silicon Graphics/Cray Origin 2000]. It's not madly expensive, maybe $4 million. Its peak speed is about 75 gigaflops, under 10 percent of a teraflop.
BG: A teraflop?
Giles: A computer capable of executing a trillion operations a second. A gigaflop is a billion. Desktop machines can do tens of millions of such operations a second.
BG: You've done pioneering work in high-end computing.
Giles: Yes, but what I'm most proud of is the work we've done here at the Center for Computational Science to make high-end computing available to a range of academic researchers around the region and country. It infinitely improves the capacity to link theory and experiment.
BG: What are the most complex kinds of calculations made on high-performance machines?
Giles: One is modeling weather and climate, which involves hundreds of millions of variables. In the microscopic realm, you get similarly big problems. I was involved in simulating patches of surface of the kind of magnetic material in disk drives, studying features on the scale of nanometers, which are billionths of a meter. The goal of the new Department of Energy program called [Advanced Strategic Computing Initiative] is to replace a lot of experimentation and testing of nuclear weapons and materials by computer simulations. That probably is the largest single supercomputing effort in the world and is driving machines on the 30-teraflop scale.
BG: Are computers getting smarter than people?
Giles: The computer is so far from being able to act or to interact autonomously in the way that we would recognize as possessing motive or anything we can do, it's hard to make that argument. They're smarter in an idiot-sa vant way - doing circumscribed tasks brilliantly.
BG: Where are supercomputers most directly affecting our lives?
Giles: In engineering and manufacturing simulations where one wants to explore the boundaries of processes where experiments are too difficult or expensive. Airplanes have been designed using virtual wind tunnels. Computerized car-crash simulations help focus auto design.
BG: Are physics and computational science a largely male domain?
Giles: Yeah, but there's no logical reason. And there are very few African-Americans as well.
BG: What are the barriers?
Giles: One is that to get into these fields, you need to be on the right track early, but for many women and minorities, the necessary math and science is either not available or they're discouraged from pursuing it. They tend to be very isolated in their departments - I didn't meet a black physicist until I was a graduate student. All too often, the experience of African-Americans and women in the fields where they're underrepresented is that they're given the message that ``You're not in the right place.''
BG: Are you addressing this? problem?
Giles: At the student level, the big thing is to give people a sense of a broader peer group. I'm directing a national program that is working to ensure women and minorities access to and inclusion in advanced computational science. A much harder piece is trying to deal with faculty promotion and tenure, where key decisions are made behind closed doors.
BG: Does the lack of some schoolchildren's access to computers aggravate race and class problems, as has been suggested?
Giles: It's going to be a problem. Boston seems to be doing a pretty good job compared to a lot of other cities. Computers are not an educational panacea, but, potentially, the onset of this information age is an opportunity for lots of people to come to the top and expand the success of their communities if they get engaged.
BG: Is there a pattern of cyber-segregation?
Giles: There's a government study that shows poorer communities are less well wired than rich communities, black less than white.
BG: Are you optimistic about the future?
Giles: I tend to be generically optimistic about everything. But it's not at all clear in terms of stuff like government policy that there's enough to hang that hope on. I'd like to see a government mechanism that would say that part of being an American citizen is having access to these technologies.
BG: Do you think the Y2K computer problem is as serious as doomsayers say?
Giles: The wildest statements - head for the hills with tin cans full of beans - are clearly off-scale. I'm not that worried about it.
12/5/1998
| This website was created by and is maintained
by Dr. Scott Williams, Professor of Mathematics State University of New York at Buffalo |
|
|
|
|
