DARPA awards high-performance computing contracts

The Defense Advanced Research Projects Agency has awarded contracts to three companies for the second phase of a program designed to create powerful new computer technologies.

Called the High Productivity Computing Systems (HPCS) program, the DARPA effort intends to create powerful, scalable and economically viable high-powered computer systems suitable for use in national security and industry by 2010.

DARPA selected Cray Inc., IBM Corp. and Sun Microsystems Inc. as the three primary contractors. The program is expected to bridge the gap between current high performance computing and quantum computing, an innovation that is still years away.

Cray, partnered with New Technology Endeavors Inc., calls its effort Cascade, and received $43.1 million from DARPA. IBM got $53.3 million for an approach it calls Productive, Easy-to-use, Reliable Computing Systems, or PERCS. Sun received $49.7 million to continue work on their integrated system approach, called Hero. High-performance computing poses a number of challenges that companies have struggled to overcome, said Mike Vildibill, director of government programs for Sun. One such challenge is the speed of data transmission across components within a computer, or through a network, he said. Just as a chain is no stronger than its weakest link, a computer is no faster than its slowest component.

"Another issue is the complexity of programming it," he said. "They're funding thousands of programmers to program these huge high end computing systems. What is the performance [gain] if it takes 10 years to write a program?"

During the first phase of the High Productivity Computing Systems program, companies submitted concept studies and plans, which DARPA then evaluated. Vildibill said the government's involvement and funding is critical to making progress.

"The problem is that the market is not a big market," he said. "It's not one to put your best people on. [However], we believe the high-end computer market is often an indicator to future commercial market requirements. It's often the case that the type of computing taking place at the very high end is exactly the same we see enter into the commercial market four or five years later."