Press Archive - 2000-Apr, Scientific American

Power to the PC

Scientific American
April, 2000
by David Pescovitz

Distributed computing over the Internet goes commercial

To date, more than 1.7 million people have participated in the largest computation in history. These scientists, students and PC hot-rodders aren't shooting for a Guinness world record, either. The aim of the SETI@home project is to discover life on other planets. Individuals volunteer to put idle Web-connected computers to work analyzing the gigabytes of raw data collected by the Arecibo Observatory in Puerto Rico. The reward for joining this massive experiment? Prestige, mostly. And the possibility of being part of first contact. But can the same distributed-computing paradigm be used to turn a profit?

Absolutely, says Adam L. Beberg, a 26-year-old computer scientist. Beberg and a handful of colleagues are cranking away on Cosm, a set of software applications, programming tools and protocols to commercialize distributed computing. "I'm trying to build an infrastructure where a company could run our software and utilize 100 percent of their resources," Beberg explains.

Beberg had helped bring this notion of distributed computing to the masses with distributed.net, which he founded in 1997. It enlisted computer users on the Net to crack encryption keys for a contest sponsored by RSA Labs. Distributed.net now boasts the processing power of 160,000 Pentium II 266 megahertz computers working all day, every day, on similar code-cracking endeavors. Other academic and research outfits are also employing on-line distributed-computing techniques, which chop data into manageable chunks of "work units" in other processing-intensive applications: to find huge Mersenne primes (those numbers following the formula 2p - 1, where p is prime), calculate the quadrillionth bit of * (and beyond) and, potentially, conduct a 21st-century climate simulation.

Beberg sees plenty of opportunity for the commercialization of distributed computing. For instance, a pharmaceutical company may want to search for a new drug via computer models of viral agents. Or a digital-animation studio might need to render 100,000 high-resolution images for a feature film. Both tasks require tremendous computational power, yet most computer laboratories are limited by the number of PCs they have direct access to. "With Cosm, you could effectively take the computers used by, for example, customer-service representatives who hit a key every few minutes, probably playing solitaire, and give that computing power to the research department," says Beberg, who is simultaneously searching for venture capital while putting the finishing touches on Cosm.

Whereas the term "distributed computing" is commonly used when referring to projects like SETI@home, Beberg points out that his more robust Cosm system is more in line with the field's 30-year history than are the examples of "collaborative computing" currently on-line. Cosm's platform-independent software will run on any computer, is outfitted with strong security features and, perhaps most important, enables the client computers to talk with one another as well as with the server. "With problems like a weather simulation, where each piece of data depends on other pieces, you need to have communication," Beberg remarks.

The next step, of course, is to let Cosm loose on the Internet. Corporations, Beberg believes, will "hire" on-line users to yield their spare processing power, compensating them with cash or gifts. And that's when the market potential for "public" distributed computing will test its legs.

Several university projects are traversing a path similar to Beberg's - notably Globus, a project of the University of Southern California's Information Sciences Institute and Argonne National Laboratory, and the University of Virginia's Legion "worldwide virtual computer," whose offshoot, Applied Meta, counts numerous research, military and academic institutions as clients. But Cosm's latest private-sector competitor is Porivo Technologies, which has just scored nearly $1 million in venture capital. More a marketing team than a cabal of young computer scientists like Cosm, Porivo hopes to purchase the core of its operating software and launch what it calls the first "computer-processing service bureau." The Porivo Web site, basically a distributed-computing portal, would contain distributed-computing projects from which users could pick and choose. "If we can identify a clinical project, say, a cure for asthma, then we can potentially build a community of people who want to help," says Porivo CEO William Holms, who notes that corporate clients will also be targeted.

If altruism isn't enough to lure participants, Holmes believes that subsidizing users' Internet service or giving them frequent-flier miles could do the trick. Ultimately, the commercialization of Internet-based distributed computing will most likely prove the old adage that you get what you pay for.