Supercomputer clears view of space

• By Dan Caterinicchia, Dan Caterinicchia
• Nov 21, 2000

Satellite, spare part or UFO?

Distinguishing objects in space has become clearer today, following the announcement that the Air Force's space surveillance team is using a new IBM Corp. supercomputer to identify the more than 9,000 objects in Earth's orbit.

The Air Force team, based at the Maui High Performance Computing Center in Hawaii, is using a supercomputer upgrade to speed the digital enhancement of anything located, tracked or imaged using ground-based telescopes, said Capt. Brian Beveridge, program manager for MHPCC and project officer for the Air Force Research Laboratory.

The IBM SP supercomputer is capable of 480 billion calculations per second and can rapidly turn fuzzy, unreadable telescopic images into sharply focused motion pictures. This enables Defense Department officials to determine whether the object is a spacecraft or some spare parts left behind during a NASA flight.

"It's the same software and algorithms as before, but it allows us to do more images in the same amount of time," Beveridge said. "It's like upgrading your desktop from a 200 MHz to a 700 MHz machine.... You can process data faster."

The supercomputer may also be used during flights of the space shuttle, which will enable NASA controllers to identify damage that might have occurred during liftoff or in flight.

The Air Force and DOD are using the supercomputing processors for other missions besides space image rendering, but those are classified, said Gene Bal, director of MHPCC.

Bal said that the facility has been using IBM SP technology since 1993, and the latest supercomputer was part of a $10 million option. This recent upgrade was valued at about $4.5 million.

The latest IBM SP has 80 nodes with four processors each and uses complex algorithms to significantly improve telescopic images in about five seconds, as opposed to 10 seconds in the past, said Dave Turek, vice president of IBM's deep computing unit. The parallel programming setup enables extremely fast communication between the nodes and the ability to solve complex problems.

Initial benchmarks with the new supercomputer show that the processing time has been cut in half for many applications, said Steve Karwoski, associate director of operations at MHPCC. "For most applications, we're getting approximately a doubling of performance," he said.