Madison, WIS. — Jill Bower reports that Condor, a relatively new computer concept that can save thousands of hours of time for a researcher, employs technology that uses other computers’ power while they are idle. Condor is a research project that has been underway in the computer science department for more than 10 years at University of Wisconsin-Madison and was started by principal investigator Miron Livny. Condor harnesses power from idle computers and uses this power to perform jobs that are considered compute-intensive.
Condor is able to run due mainly to the fact that most computers on campus, throughout any 24-hour period, are idle about 60 to 70 percent of the time, said Condor program manager Todd Tannenbaum.
“If you’re doing anything compute-intensive that you need to run on a computer, you can submit your job to Condor,” Tannenbaum said. “Condor can then look out over your network and find machines in the network that would otherwise be idle. It can harvest all your otherwise wasted and idle resources.”
According to Tannenbaum, a scientist might submit many stimulation runs to Condor, and Condor will run these on machines throughout the organization and migrate the job from one machine to another as they become available.
In the past month, Condor delivered more than a quarter of a million hours of compute time to researchers using the computers in the computer science lab alone.
“This means that if you have just one fast computer, you would need to run that computer for 257,000 hours to get the amount of stimulation work done that Condor delivered to the researchers,” Tannenbaum said. “In order for the computer to do all the work that Condor delivered to it in the past month, it would have to run for 30 years.”
The Condor program is now being used at UW by various researchers. There are approximately 1,200 to 1,300 computers that are running the Condor program currently, Tannenbaum said. Two hundred organizations have installed Condor, including other academic institutions, government institutions such as NASA and commercial companies.
Some applications of Condor are being utilized at UW to help solve intensive problems. Michael Ferris, a computer science professor whose main focus is solving optimization problems, has been using the Condor program for the past three or four years. “I work with optimization problems that either maximize profit or minimize cost,” Ferris said.
“Optimization problems appear all over the place. I’ve been looking at some large-scale problems that haven’t been solved by any other method.” These problems are now solvable due to the large amounts of power that Condor provides, Ferris said. “We’re able to use large amounts of computing time with Condor,” Ferris said. “There was a problem that hasn’t been solved for the past 12 or 13 years. By using Condor, which provided 9,000 hours, we were able to solve this problem.”
Condor can also manage what are now being called “commodity compute clusters.” “Instead of buying a really expensive supercomputer, people are buying 100 regular PCs, sticking them on shelves, and putting software on them to make the 100 computers work together as one big, fast computer. Condor is one of the software components that can help you do that,” Tannenbaum said.
But Condor’s best applications may be yet to come. The program is expected to be highly effective on projects such as UW’s genome mapping work and NASA’s simulation projects that deal in space technology.
“We have a great deal of evidence now showing that enabling people with computing power acts as an intellectual fertilizer,” Livny said. “If you give them power in a form they can use, people can do wonderful things.”