Scientific Computing for the Masses

End of January 2008

The temple of scientific computing has been opened to unwashed heathens. Science has always pushed the envelope of computation and the state of the art of computers. Until a decade ago, supercomputers built by IBM and Hitachi and Cray duked it out to claim the title of fastest calculator; their speeds were measured in teraflops, where a teraflop is a mind-bending trillion floating point operations per second. Such machines were power-hungry and cost tens of millions of dollars and they had sleek housings made of anodized aluminum. It took chilled water or Freon to sooth their fevered circuits. Supercomputers were cool, literally and metaphorically.

Supercomputers still exist and they’re used to tackle the toughest problems in science: how galaxies form, how proteins fold, what happens when an atomic bomb goes off underground, and how to embarrass mere mortals at the game of chess. The current record-holder is an IBM machine called Blue Gene that clocks in at a staggering 500 trillion operations per second. It could do you taxes in a nanosecond, if it would ever stoop that low.

But the relentless pace of Moore’s law means that a high end desktop today is like the supercomputer of only a decade ago. Moreover, blazing speed in a supercomputer is achieved by clustering processors and running them in parallel. As PCs gained speed researchers realized they could be harnessed into a highly distributed supercomputer, where the calculations are parceled out over the Internet and the answer is assembled afterwards.

Rather than a supercomputer in a room, this is a supercomputer in thousands of rooms scattered across the world. There are a half a billion PCs around the world. They spend most of their time idling, and rarely approach their computational capacity even when they are being used by their owners. Scientists have figured out how to harness this vast excess capacity on a volunteer basis. Most people are delighted to be able to help solve problems at the cutting edge of science.

Volunteer or networked computing has projects across all scientific disciplines, spurred by an open-source platform to make it easy to set up and manage these projects called, with onomatopoeic whimsy, BOINC, or the Berkeley Open Infrastructure Initiative for Network Computing. Projects include searches for prime numbers and cancer cures and testing for drugs that can combat tropical diseases. IBM has harnessed 800,000 volunteer computers for a variety of philanthropic and humanitarian causes. A group from Stanford studies protein-folding, which is the key to a variety of diseases such as Alzheimer’s, and late last year they used the processing power of 40,000 Playstation 3’s to pass the magical petaflop barrier, a quadrillion floating point operations per second.

Astronomy features heavily in this new form of “citizen” computing. The granddaddy of distributed computing projects is SETI@home, which started in 1999 and has had over 6 million participants. This screensaver analyzes chunks of radio spectral data looking for artificial signals from extraterrestrial civilizations; none has been found so far. Some of the more interesting distributed computing projects tap the skill and judgment of their participants, not just their CPU’s. Stardust@home enlisted 24,000 volunteers to search images of the porous aerogel from the Stardust comet sample return mission for telltale tracks of interstellar dust grains. This needle in a haystack project was very successful; forty million searches resulted in fifty new dust particles recovered.

The Galaxy Zoo has been even more of a hit. This British-American collaboration seeks civilian scientists to classify the shapes and types of galaxies from Sloan Digital Sky Survey images, of which there are too many for professionals to handle. In under a year, 100,000 volunteers have classified over a million galaxies. Project managers had to buy new servers to handle the overwhelming public response. Each galaxy is classified by as many as 30 volunteers. The results are just as accurate as classification by a professional.

It’s unabashed good news that computing power is making science less of a priesthood, and letting members of the public participate in cutting-edge research or help find the answers to real-world problems. If you are not doing research right now, you’re almost certainly wasting the computer in front of you.

Comments are closed.