Web Guide Turns Playstation 3 Consoles into Supercomputing Cluster
It's no secret that Playstation 3 (PS3) hardware has enough compute power to run a supercomputer. Harnessing that power is so easy that a university team has posted instructions for do-it-yourselfers to build their own souped-up consoles.
Drawing from their own experiences using PS3s for gravity research and pattern recognition, University of Massachusetts Dartmouth professor Gaurav Khanna and principal investigator Chris Poulin set up a Web site (www.ps3cluster.org) that provides step-by-step details on turning gaming consoles into nodes for a supercomputing cluster. The Linux-based nodes would provide an affordable alternative to renting time on actual supercomputers, the researchers said.
"Science budgets have been significantly dropping over the last decade," Khanna said. "Here's a way that people can do science projects less expensively. This new Web site will show people how to move forward."
According to the guide, setting up a PS3 as a cluster node requires only a few steps—from setting up a version of Fedora on the device to installing the message passing interface (MPI) that enables nodes to communicate. Everything needed to set up the cluster is open source—including the guide—and the only extra hardware required is a USB memory stick and a PC to burn a couple of discs.
The researchers, who operate their Gravity Grid cluster with financial backing from the US National Science Foundation, hope the guide will make clustered computing easier and eventually provide a framework for high-level grid development tools. "This opens up a huge door to partnerships with industry and other universities," Khanna said.
The scientific community began looking at Sony's game consoles as supercomputing devices in 2003, when the only available model was the Playstation 2. At that time, researchers constructed supercomputing clusters by using the machines' graphics co-processors, called Emotion Engine, which could execute up to 6.5 billion mathematical computations per second. The National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign famously built a cluster out of 70 PS2s for US$50,000.
The next generation of Playstation consoles contained even more raw power with the inclusion of the Cell processors, developed by an alliance of Sony, Toshiba, and IBM known as STI. Khanna estimates that Cells—which contain a main power processing element (PPE) and eight synergistic processing elements (SPE)—run faster than desktop workstation chipsets and offer better computing power per dollar than anything else on the market. Adding to the Cell's allure is a specialized variant used in the IBM Roadrunner supercomputer, currently the fastest in the world.
Scientists quickly began using PS3 for their supercomputing power shortly after the machines debuted in late 2006. Stanford University's Folding@home project, a large distributed computing project that borrows users' home computing power to measure protein folding, made it easy for PS3 owners to contribute and produced a YouTube video for extra publicity. According to the Folding@home Web site, the addition of 50,000 PS3s to the project helped it reach 20 gigaflops per computer.
Khanna started his Gravity Grid project in 2007 after writing code to optimize the Cell's performance, and now uses 16 PS3 nodes to measure gravitational waves created by the merger of two black holes.
Share this article