Google takes on supercomputing

The Mountain View, Calif.-based Internet search company invited 500 people to try out a new version of its toolbar that lets Windows users donate their computers' otherwise unused processing power to the Folding@home project at Stanford University. The project seeks to figure out how genetic information is converted into proteins, complex molecules whose three-dimensional structure is key to everything from fighting off a cold to transporting oxygen around the body.

The work is the latest example of the distributed computing movement, in which computing jobs are farmed out in small chunks to ordinary PCs across the Internet, finding a use for otherwise untapped processing cycles. The movement has had grand ambitions to cure cancer, but thus far its chief successes have been curiosities such as the discovery of gigantic prime numbers.

There's no denying the popular appeal of some of the projects, however, which can pit hundreds of thousands of participants in contests to see who can crunch the most numbers.

The Google Compute project illustrates how the approach to even the most ornery problems of computing science is changing. Supercomputers once were isolated, expensive systems affordable only to the likes of aerospace companies, national laboratories and well-funded universities. But all that is changing with the arrival of the Internet, omnipresent PCs and ever-faster network technology.

"The main motivations were to try to leverage Google's expertise with large computer systems and to try to give something back to science," said Susan Wojcicki, a Google product management director and the head of the Google Compute project.

Google co-founder Sergey Brin initiated the project, Wojcicki said, and people started trying the software two weeks ago. An option on Google's toolbar lets the participants in the project download the necessary software to their computer. Google is considering offering the program to a larger audience, Wojcicki added.

"From what I saw, it simply rocks!" said one enthusiastic person who sampled the software. "When I move my mouse across that little DNA icon, it tells me what protein it is folding and what percentage it has completed."

A moneymaker?
Google likely will expand the program to include other scientific endeavors, and possibly even computational problems, to benefit its search business, Wojcicki said. But Google Compute isn't likely to become a source of revenue.

"You never want to say never, but the goal now is to contribute something to science. We have enough fish to fry in our current businesses," Wojcicki said.

Google, having secured its position as a top search engine, has been pulling out all the stops to increase its revenue--likely in anticipation of an initial public offering, some believe. The company's grander aspirations have been visible in features such as a news feed, targeted advertisements, commercial search services and catalogs.

Distributed computing is just one part of the overhaul of the supercomputing world. For one, existing supercomputers are being linked into "grids" of shared computing and storage resources such as the Energy Department's Science Grid, unveiled Friday. For another, groups of inexpensive Linux computers can be linked with high-speed networks to form a cheap "Beowulf" computer.

Yet distributed computing has given supercomputing efforts popular appeal.

The best-known distributed computing project has been SETI@home, which scans radio telescope signals for extraterrestrial communication signals. Distributed computing began with more abstruse projects, however, such as hunts for Mersenne prime numbers, optimal Golomb rulers and Fermat numbers.

Though SETI@home hasn't uncovered any alien chitchat, distributed computing has had its successes. Most recently, one effort with 210,000 participating computers uncovered a 4,053,946-digit prime number, the largest found so far.

Getting the word out
Google's toolbar addresses one of the key obstacles in distributed computing: propagating the software to all the computers involved in the effort. And if people are eager to participate in the distributed computing program, they might be more inclined to install the toolbar, which beefs up Web browsers with links to Google's search engine.

The Google Compute software works on Windows 98, Me, 2000 and XP, Google said.

In the more exuberant climate of Internet business of the late 1990s, several start-ups seized the idea that money could be made by selling processing power to pharmaceutical companies and others.

The prospect faded, though; one site, Popular Power, shut down, and e-mail provider Juno Online Services ran into controversy when it tried out the idea. Distributed computing as a business prospect today generally is focused on using a corporation's own computers, a much more controlled and predictable environment than the entire Internet.

Companies involved in this arena included United Devices, Turbolinux, Sun Microsystems, Parabon Computation, Platform Computing and Avaki, formerly Applied MetaComputing.

Some companies still sponsor distributed computing projects that extend to the entire Internet, however. United Devices' network helped to screen molecules that could be related to anthrax. And Parabon is involved with the Compute Against Cancer effort to boost cancer research.