Supercomputing in space
Many of you know that the computers sent into space are not very powerful compared to the ones we use everyday on our desks. This is because space is a pretty tough environment: computers need to be ruggedized because of cosmic radiation, and this slows their performance while increasing their weight and cost. Now, engineers at the University of Florida (UF) and Honeywell Aerospace are designing the first space fault-tolerant supercomputer. It should have 20 processors for a computing power of about 100 gigaflops. If everything goes well, this first supercomputer will fly aboard the unmanned ST8 rocket mission in February 2009. But read more...
So why this new high-performance computer will be that powerful? The idea is simple: the computers will not be hardened but will be fault-tolerant. If one fails because of radiation, another one will take the control.
The UF-Honeywell computer aims to upgrade both satellites and probes with a novel design called the Dependable Multiprocessor. Funded by NASA's New Millennium Program and the Florida High Technology Corridor Council, the goal is to cope with radiation from solar flares or other space events not through the physical hardening of components -- but rather through software that allows the computer to survive radiation-caused flaws or errors.
Below is a diagram showing the reconfigurable cluster computer which will move into space in 2009 (Credit: NASAs New Millennium Program)
And how will work this "Dependable Multiprocessor for Space"?
Alan George and his team of graduate students develop and evaluate concepts and elements of the system. As per the project’s requirements, they feature off-the-shelf components with no deliberate radiation hardening. Their methods involve strategies such as making the computer fault-tolerant, or able to make an instant switch from a temporarily failing board to a functioning one. They also use algorithm-based techniques to detect and correct processing errors. “If one board is failing because of radiation, we can automatically go to another,” George said.
But will such a computer cluster really work in space?
Even with the radiation problem solved, that’s a huge challenge because the system must be small, lightweight, capable of surviving the vibration of launch and the shock of the delivery vehicle separating from the booster rocket –and operate on relatively little precious electricity, among other challenges.
For more information about this project, you should read a presentation made by John Samson, of Honeywell Defense & Space Systems, about the NMP ST8 High Performance Dependable Multiprocessor (Microsoft PowerPoint format, 33 slides, 5.09 MB, September 20, 2006). It contains tons of details about the specifications and it's a must-read paper.
And for more scientific details, you should read some of the papers published by the UF's Advanced Space Computing group. Several of these papers are not even published but are available on the UF site and are also worth reading.
Sources: Aaron Hoover, University of Florida News, October 26, 2006; and various websites
You'll find related stories by following the links below.