Computer World –
In the Microsoft Corp. antitrust case, the government is trying to bring diversity to the PC operating system market through litigation. But one federal agency is using money rather than legal action to help ensure that there's competition in an area of computing important to academic and industrial research: supercomputing.
When the National Science Foundation (NSF), an independent federal agency in Arlington, Va., selected Compaq Computer Corp. earlier this month to build its most powerful supercomputer to date, it said it was picking a company that put together the best proposal for the job. But the NSF said that by choosing Compaq, it was also promoting diversity in supercomputing applications.
It's a funding approach that may ultimately help companies seeking more computer power for advanced industrial and research applications, experts said.
"You want some diversity and competition in the program. . . . You don't want everything stacked up in one place," said Bob Borchers, director of the NSF's division of advanced computational infrastructure and research. "On the other hand, if you have too many mouths to feed, it's very, very hard to keep them all at the state of the art."
Compaq will be building a 6-teraflop (1 trillion calculations per second) computer at the Pittsburgh Supercomputing Center under a $45 million project approved by the NSF, which said it may well be the fastest supercomputer for nonmilitary uses.
At two other supercomputing centers used in research -- the San Diego Supercomputer Center and the National Center for Supercomputing Applications at Urbana, Ill. -- different systems are in use. San Diego has an IBM machine operating at about 1 teraflop, and Urbana has a Silicon Graphics Inc. system operating at less than 1 teraflop, said Borchers.
"I think there is always a benefit for the government to having competition in the marketplace," said Tom Tecco, director of global computer-aided engineering test systems at General Motors Corp. in Detroit. And those gains aren't just in the price and performance of supercomputers but in the development of different views on how the technology should evolve, he said.
The supercomputing projects' main contribution to industry is in their use of pioneering technology, said Tecco. "Government-sponsored projects have helped to shape the industry in its early stages," he said.
For instance, GM is working on computer applications to develop a "digital car" -- a vehicle that can be represented mathematically, thus allowing the company to run computer simulations of the impact on various design and performance changes. But these kinds of simulations require applications that have enormous computer capability, such as those used in studying fluid dynamics, said Tecco.
"If they can show how to do new science on this type of machine, it will be a major step forward in applying supercomputers to industrial problems," said Earl Joseph, research director of high-performance systems at International Data Corp. in Framingham, Mass.
NSF is trying to sustain Moore's Law (the 1965 prediction made by Intel Corp. co-founder Gordon Moore that computer power will double every 18 months) and wants to sponsor development of a 12-teraflop computing system next year and a 20-teraflop system the year after, said Borchers. Funding for those systems will depend on Congress.
The system at the Pittsburgh center will be an interconnected network of 682 Compaq AlphaServers, each of which will contain four Alpha processors.