Ever wonder how much you might get done if you could snatch bits of unused energy from all over the world and use them for yourself? John Grefenstette has. In fact, the new director of Parabon Labs is making a living at it.
Formerly director of George Mason University’s computational biology laboratory, Grefenstette first trained as a computer scientist when he originally started out at Vanderbilt. His interests always lay in parallelism, biology, and artificial intelligence; he developed evolutionary, or genetic, algorithms. And the more he did, the more he wanted to learn “about the biological underpinnings of these algorithms.” He went to George Mason, where “it became very clear to me that the future involved massively parallel applications to solve some of the problems in computational biology,” he says.
Parabon’s business of distributed computing is built around such parallelism. The company’s technology takes underutilized computers connected to a network and uses their capacity to work on problems that are too big for other machines. Users can distribute their problems to computers connected to the Internet or specify distribution to unused computers in their organization. “I heard a little bit about the Parabon approach,” Grefenstette says, and he decided to get involved.
“My role here is to prioritize the most important applications and interact with computational scientists in industry and government to see what the most promising directions are.” Another aspect of Grefenstette’s job is to increase the visibility of distributed computing.
Grefenstette believes one of the best parts of Parabon’s technology is that it’s inherently self-evolving: as various people upgrade their computers, the whole system gets faster, at little or no cost to Parabon customers. But by no means is the technology going to wipe out supercomputers or clusters. “We see this as complementary to existing solutions,” he says. “We’re not going to make compute farms obsolete.” But with distributing computing, Grefenstette says, users “get the effect of having a second compute farm.”
— Meredith Salisbury