In brief
By Staff -- Manufacturing Business Technology, 9/1/2003 12:00:00 AM
Supercomputer tests could improve enterprise system reliability
The term supercomputer conjures up images of large mainframes grinding through mammoth calculations in isolated, refrigerated rooms. In reality, today's supercomputers are made of large clusters—as many as 100,000 to 200,000—of the same microprocessors that run desktop computers and network servers. Supercomputers do, however, work on complex problems, meaning the people running them must have ways of predicting when the computers might fail. Hewlett-Packard's research labs are working on such methods, hoping to pass that knowledge to developers of HP's enterprise servers. "The aim is to take techniques that work with 20,000—maybe even 200,000—processors and make them available to all of HP's enterprise computing users," says Richard Taylor, a researcher at HP Labs.
