Linux clusters built from Facebook's blueprints will help crunch numbers for the US government's hydrogen bomb scientists.
The computer system, dubbed the Tundra Extreme Scale series, will cost $39m, and at its peak perform between seven and nine thousand trillion math calculations per second – that's seven to nine petaflops.
The machines will be installed at America's Los Alamos, Sandia, and Lawrence Livermore national laboratories from April 2016, with the last rack scheduled to be in place by September 2018. There, they will carry out "stockpile stewardship," which is a wonderfully sterile and bureaucratic way of saying nuclear weapon reliability testing and simulation.
Essentially, the computer system will be used to calculate whether or not Uncle Sam's stockpile of nukes, stored away in grim silence, can be relied upon to wipe cities from the face of the Earth at short notice. Discovering your thermonuclear warheads have deteriorated into duds only after you press the big red button will be a bit of a bother. Politicians and military commanders want to avoid that scenario.
The Tundra series will be powered by Intel Xeon E5-2695 v4 processors, and built from servers made by Silicon Valley's Penguin Computing using Open Compute Project designs.
The Open Compute Project (OCP) was founded in 2011 by Facebook when the social network's engineers effectively open sourced the blueprints for their customized servers and racks. The Tundra series will be one of the world's largest OCP installations when it's fully powered up.
Penguin will deliver the hardware in so-called scalable units that each provide 200 teraflops in computing power. These units will be clustered together and distributed across the three labs.
Doug Wade, head of simulation and computing at America's National Nuclear Security Administration (NNSA), said in a statement: “These clusters will provide needed computing capacity for NNSA’s day-to-day work at the three labs managing the nation’s nuclear deterrent.
"This tri-lab effort will help reduce costs, increase operational efficiencies, and facilitate collaborations that benefit our nation’s security, support academia, and advance the technology that promotes American economic competitiveness.” ®