PCs of the World, Unite!

Startup Entropia, and others, hope to solve vexing problems by pooling millions of underused computers into powerful networks

If you're a computer user, chances are your machine is standing idle most of the time while you're yakking on the phone, grabbing a latte, or schmoozing at the water cooler. And that's a shame, because all that unused computing power is a tremendous resource that charities and businesses would love to tap into. Using a technology called distributed computing, all those spare CPU cycles on your PC can be channeled into "supercomputing" networks that can be set to work on hugely complex data-analysis problems.

FightAIDS@Home, for example, which was launched in September and has already signed up 15,000 computers in 80 countries, is searching for a cure for AIDS. And SETI@home, a project based at the University of California at Berkeley, is trying to discern signs of extraterrestrial intelligence in radio signals from space. But distributed supercomputing is also starting to be used for a purpose pertinent to business.


  On Mar. 21, Entropia Inc., a San Diego distributed-computing-services company, will launch www.safermarkets.org, a public-service project whose ambitious goal is to accurately predict volatility in financial markets. "We're hoping for tens of thousands of members" on the network, says Entropia Project Director Tim Cusac. "But we can start doing the research as soon as one computer signs up."

Although Entropia is focusing mainly on nonprofit efforts, it does hope to make money by charging some corporate clients to use its network. That's already becoming a crowded field, with half a dozen competitors trying to build for-profit distributed supercomputers. The largest will likely be free Internet service provider Juno, which hopes to launch its Virtual Supercomputer Project (VSP) in the next few months. The ISP aims to sell the spare CPU cycles of its 4 million subscribers to pharmaceutical- and medical-research companies, just like Entropia. Juno figures that it can collect willing volunteers from the 80% of its subscribers who pay no fees. But if those volunteers don't materialize, Juno may require participation in VSP in exchange for the free service.

But Entropia, with its feel-good mission, might prove the strongest draw. Signing up is simple: Participants log on to the Web site and download Entropia's software. That's it. As long as the Internet connection is on, the computer takes care of the rest -- contacting the server in San Diego for data-analysis tasks, performing the calculations, and then feeding the results back silently in the background, behind the donors' own work. Those unused CPU cycles will be utilized whenever the machine is on -- even in the milliseconds between the owner's keystrokes. "It's really cool because in one one-hundredth of a second you can rescue those CPU cycles that aren't being used," Cusac says.


  The resulting networks are very powerful. Entropia is currently hosting FightAIDS and another project, called Mersenne, which is looking for the largest possible prime number (2 million digits at the last count). The power of Entropia's network is equal to 1.6 terraflops, with a single terraflop in turn able to perform 1012 calculations per second, Cusac says. That's power equivalent to the world's largest IBM supercomputers -- Asci Blue Pacific at Lawrence Livermore National Laboratory and Asci Blue Mountain at Los Alamos National Laboratory.

The objective of the volatility project is to nail down a formula that can predict the likelihood, degree, and duration of volatility in the Nasdaq and S&P indexes and in five currency exchanges where the U.S. dollar is half the equation. Entropia's partners in the project are business-school professors Michael Brandt, with the University of Pennsylvania's Wharton School, and Christopher Jones of the Simon School of Business at the University of Rochester in New York.

The project's first phase, scheduled to last three to six months, will attempt to create and optimize a formula, using Bayesian statistics (which deal with human behavior) to generate a random fictional history of volatility. During phase two, lasting six months, the network will fine-tune that formula against real historical data. Phase three -- still being planned -- will apply the formula to individual equities in order to predict their volatility.


  The final results will be published in economic journals and made available to the public for free. The point of such "cause computing" is to help people improve their finances through better planning tools. "The goal of the project is to understand volatility better, which means that if I'm going to retire in three years, [it] will help me decide whether to invest more heavily in equity or fixed income," Cusac says. "Basically, this is about asset allocation." But the data it generates could also be helpful to market professionals such as derivatives traders.

Entropia's future also may be riding on the project. While the company dabbles in good causes, it's making money, too, by routing about 10% of those donated CPU cycles -- with users' full knowledge -- to commercial clients doing life-science and Web-testing computations of their own.

For their part, Entropia personnel view their effort as equal parts altruism and capitalism. Cusac says the company's engineers take pride in donating their technology to solving some of humanity's greatest conundrums. "This is a powerful form of computing," he says, "and we feel strongly that we have the responsibility to do the right thing with it, whether that be in the philanthropic realm or the commercial realm." After all, this is a project that won't just generate revenues. If all goes well, it could also change the world.

By Joan Oleck in New York

Edited by Thane Peterson

    Before it's here, it's on the Bloomberg Terminal.