Floridians to design first space-based supercomputer

Hey, Dave, what are you doing?

By Lucy Sherriff • Get more from this author

Posted in Servers, 31st October 2006 11:46 GMT

Free whitepaper – Selecting an Industry-Standard Metric for Data Center Efficiency

Engineers and researchers at the University of Florida are hard at work building a supercomputer capable of surviving in orbit. The computer is scheduled for a 2009 launch, on board a NASA test mission, and will be 100 times more powerful than anything already orbiting Earth.

So far, orbiting computers have missed out on decades of technological advances because of the need to "harden" anything electronic to protect it from cosmic radiation. This makes things slower and bulkier, not great when weight is such a critical issue.

However, as the amount of science being done by orbiting satellites increases, so does the need for more powerful number crunchers above the atmosphere, and faster downlinks to Earth-based researchers.

"There are only so many bits per second you can send down from a satellite," said John Samson, the principal investigator for the project at Honeywell's Clearwater facility. "That means scientists are very limited in how much science they can do."

The technology developed will also be needed to build more autonomous space craft that will be more capable of making course corrections, for example.

"To explore space and to support Earth and space science, there is a great need for much more processing power in space," said Alan George, a professor of electrical and computer engineering at the University of Florida (UF). "To be autonomous is to require a lot of computation and, until now, conventional space processing technologies have been incapable of high-performance computing."

The UF researchers are not trying to find lighter and smaller ways of hardening it, instead they are trying to find ways of building a a machine that is capable of functioning in a radiation intensive environment. Techniques include making a machine especially fault tolerant, being able to switch from a failing circuit board to a functioning one, and using algorithms to check for errors. ®

Free whitepaper – Guidelines for specification of data center power density