News

P2P By Any Other Nam

You can't open a newspaper or download a tech-news Web site these days without seeing "Napster" splashed all over the place. There's a reason Napster keeps getting all the peer-to-peer (P2P) computing headlines: the music-file-sharing service has signed up more than 58 million users in the two years since it was launched by an 18-year-old college student, and it continues to register more than 300,000 people every day. Throw in a couple of big lawsuits from music industry moguls who want to shut the service down, and it becomes a veritable headline machine.

But all the Napster noise is stealing the thunder away from what may be the more important P2P model; to differentiate, call it "distributed computing." Where Napster-like systems support file-sharing among individual PCs interacting directly with each other with little or no server involvement, distributed computing architectures organize linked machines for the purpose of sharing computing cycles. The first approach is about people interacting; the second is about marshalling computing power-and in the long run, it's the second one that has the most potential to influence a company's bottom line.

Computers running software from companies like San Diego-based Entropia utilize idle computer time drawn from a network of PCs to solve large computational problems, run financial analyses, and find cures for diseases. Working in the background, these programs take on small workloads that have been divided and distributed to various devices across the Internet--essentially borrowing processing power from other computers on the network. This process is also called "grid computing."

"Through the use of these resources, your computing device can be both a consumer and a producer on what becomes a universal platform," explains Scott Kurowski, Entropia's founder. "In as little as five years, we fully expect end users to be running applications that invisibly tap into huge amounts of computing capacity on the backend.

Researchers at the University of Wisconsin estimate that most companies use less than 25% of their computing and storage capacity. By some estimates, 90 percent of America's computer capacity is idle at some point during the day, especially when people turn off the machine and go home for the evening. Entropia's software "rescues" these wasted computing cycles between keystrokes while applications are running on the PC.

Entropia's network encompasses nearly 100 countries around the world, from which it aggregates 50 teraflops of potential computing capacity. In fact, Entropia squeezes so many MIPS out of its network that it can afford to give away the extra computing power to good causes, such as the AIDS@home project. "This is a good way not to have to buy a lot of servers to get massive amounts of computing resources," Kurowski says.

Probably the best-known example of the distributed computing model is the SETI@Home project. Under the auspices of the SETI organization (the Search for Extraterrestrial Intelligence), SETI@Home marshals unused computing cycles of a global network of volunteer PC owners to search for radio signals from an extraterrestrial intelligence. To date, the project has attracted more than two million participants.

The technology behind distributed computing is not new. Scientists have linked supercomputers this way for years. Companies have distributed-computing software on internal networks to get extra work done at night and on weekends. Some argue that distributed computing shouldn't even be called peer-to-peer.

"I find this a bit of a stretch," says Dwight Davis. "I know that they're sharing resources among desktop machines, but to me, that's not peer-to-peer. That's really just taking the model of a massively parallel processor and extending it out to the broader universe of PCs. But the PCs themselves are not communicating with each other; it's all orchestrated by a central server, in other words a controlling authority. For me at least, P2P has to have some element of direct communication between individual computers."

Nevertheless, distributed computing companies make up a significant membership in the Peer-to-Peer Working Group; Entropia was a founding member. And the model does something else: while companies like Oracle and Sun are touting centralized computing models ("the network is the computer"), distributed computing schemes suddenly make the desktop PC relevant again. Expect to see Intel and Microsoft emerge as big supporters of this kind of P2P computing--if the Napster racket ever quiets down.

About the Author

John K. Waters is a freelance writer based in Silicon Valley. He can be reached at [email protected].