Last Month, Sandia Labs decommissioned the Red Storm supercomputer. The system was in production for 10 years and had an important impact on the HPC community. Originally designed by Jim Tomkins and built by Cray, it spawned 124 descendants at 70 HPC sites, representing more than a billion dollars worth of business to the company. Sandia Lab News covered Red Storm’s decommissioning and highlighted the bumpy road to its creation.
Cray CEO Pete Ungaro spoke with Lab News, explaining how the project saved his company. “Without Red Storm I wouldn’t be here in front of you today. Virtually everything we do at Cray — each of our three business units — comes from Red Storm,” he said.
At the time, the company was building systems with custom processors when Sandia came to them with a completely different model. Rob Leland, who was senior manager of the Red Storm project, remembers a tense design process.
“Imagine going to a company whose defining idea is one thing [a custom vectorized, or linear, processor] and telling them it’s the wrong idea for the future and they needed to focus instead on building massively parallel systems out of commodity processors.” said Leland. He believes that Cray only went along with the idea because they were on the verge of insolvency.
Some of the revolutionary design elements included commodity hardware, in particular, AMD CPUs, to achieve lower cost. The interconnect was the only proprietary part of the system. It was cooled by air instead of water, so upgrades and maintenance could be performed without shutting down the machine. Red Storm also exhibited a very upgradeable architecture. The system performed 41.47 peak teraflops in 2005 and was updated twice, resulting in a 284.16 peak teraflops by 2008.
One of Red Storm’s more secretive and now declassified achievements was known as operation “Burnt Frost.” The program was designed to model scenarios in which the system would program a rocket to shoot down a malfunctioning satellite. In 2008, the supercomputer did just that, firing a rocket from the USS Lake Eerie and destroyed a satellite travelling at 17,000 miles per hour, 153 miles above the Earth. Had the object entered our atmosphere, it could have released its toxic fuel supply.
While the endeavor to build Red Storm was risky, it resulted in enormous success for Sandia Labs, Cray and the supercomputing industry. Some believe it even spurred Intel to re-focus on the HPC market.
The system was completed in a timeframe of 18 to 22 months and Cray grew from 6 to 20 percent market share in the four years following the program. Rob Leland seemed to summarize the project best:
“Sometimes the need and vision and the people and the resources all come together to make something extraordinary happen. In my view, that’s what Red Storm represents.”