Trillions and trillions of computations

The first components of one of the world's fastest computers, the terascale supercomputer funded by the National Science Foundation, should be operational by February 2001.

NSF began the supercomputing program in the 1980s to give academic scientists the same level of computing that's available at the national laboratories, said Robert Borchers, director of the Division of Advanced Computational Infrastructure and Research in the Computer and Information Science and Engineering Directorate at NSF.

Interest and demand for computing resources has grown so much that the agency has decided to solicit more computing resources. The terascale system will be awarded for the ability to make computing power, or cycles, available to others.

"We are basically asking someone to build a power plant for us," Borchers said.

The president's fiscal 2001 budget request includes $45 million to double the capacity of NSF's terascale computing systems. The agency received $36 million in 2000 that will be awarded through a competition to a single successful vendor. Bids are due April 3, and an award is expected this summer.

Because NSF's computing resources are typically severely oversubscribed with requests, the additional $45 million in fiscal 2001 would be used to expand the capacity of the system currently being contracted and to add a second site.

The completed system will perform up to 5 trillion operations per second. It will be part of the resources provided by the NSF's Partnerships for Advanced Computational Infrastructure program and will supplement the capabilities available through the PACI partnerships with the University of Illinois and University of San Diego.

Who needs all that computing power?

* Meteorologists. Simulating a year of climate change would take a day of computer time on a typical high-powered processor. With a teraflop machine, you can do that computation in about an hour, Borchers said. Those computations will help meteorologists make better weather predictions and understand climate change.

* Biologists. Faster computing would help them predict the folding of protein molecules. That research is being used to understand the makeup of DNA.

"The software is there to do the simulations," Borchers said. "The question is, do you have the computer power to do realistic problems?"