Parallel Computing Proves to be 'Embarrassingly' Easy

Topic the subject of cover article in the American Journal of Physics

WOOSTER, Ohio - Long after classes have ended for the day, rows of computers sit idle in a darkened lab on the second floor of Taylor Hall at The College of Wooster. Students and faculty have been gone for hours, but even without a human presence, these seemingly inactive computers are actually still working - in unison - to solve complex scientific problems.

This collaborative effort is a process known as parallel computation, and it is the subject of a cover article in the April/May issue of the American Journal of Physics, which is authored, in part, by John Lindner, professor of physics at Wooster, and several of his students. The article, titled "Invitation to Embarrassingly Parallel Computing," addresses what the authors describe as the "richness and diversity of interesting physical systems that are easily parallelizable" - so easy, in fact, that it's almost embarrassing. According to Lindner and his colleagues, no special infrastructure, no sophisticated parallel algorithms, and no message-passing interface are necessary to make it work.

The system at Wooster is run through Mac OS X computers, which support Xgrid - the first distributed computing architecture built into a desktop operating system. "It's as simple as going under system preferences and checking the Xgrid box on each computer," says Lindner. "One by one, the computers are added to a cluster." This cluster then works together on a particular problem. "It's better and far less expensive than a super computer that will eventually become obsolete and costly to replace," says Lindner.

There are 31 Macs in the computer lab at Wooster, and each one has a dual processor, which gives Lindner and his students 62 processors in the cluster - enough to tackle a range of projects.

Three projects are outlined in the article, and although they are not related, each one breaks a problem into small parts and distributes those parts across many computers that work concurrently to find a solution.

In the first project, Lindner worked with Barbara Breen, the Luce Professor of Physics at the University of Portland, and Chris Weidert, one of Breen's students. Together, the three characterized the propagation of solitary ways or solitons in noisy arrays of one-way coupled oscillators. "Say, for example, you had a ball, or some other mass, connected to a spring, like the ones you find in mattresses," explains Lindner. "Normally, movement of the ball on one end would cause the ball at the other end to move, but in one-way coupling, we're looking at a way in which one ball moves and the other does not."

In the second project, Lindner worked with recent Wooster graduate Lisa May Walker and Kenyon College sophomore Kasey Kelly in an effort to visualize the space time in the vicinity of rotating black holes. Kelly extended Walker's Senior Independent Study project as part of the Wooster Physics summer Research Experience for Undergraduates program. "Black holes are so dense that light cannot escape," explains Lindner, "but if we put it in front of a background, it will deflect the rays and distort the image."

In the third project, Lindner and Wooster junior Evan Heidtmann studied solar escape as a restricted three-body problem. Their work is illustrated on the cover image of the publication, a strikingly colorful scientific visualization of order and chaos in the famous classical three-body problem, which consists of three bodies (say Sun, Earth, and a tiny projectile) interacting gravitationally according to Newton's laws. "We created an initial velocity space plot where each point corresponds to a different initial condition for the projectile launched from Earth and is colored according to how far the projectile has traveled after 50 years," explains Lindner. "For some initial velocities, the colors vary smoothly, indicating that the 50-year outcome doesn't depend sensitively on the initial conditions. For other initial velocities, the colors vary extremely rapidly, indicating that the outcome is very sensitive to the initial conditions - the hallmark of chaos."

Linder and his fellow authors are excited about sharing their findings about the power of parallel computing. "When an otherwise idle room of computers becomes a tool for teaching and learning, it is invigorating for both faculty and students," write the authors in their concluding remarks. "When a new computer cluster enables teachers and students to collaborate on cutting-edge research, it is exciting for them and their department. Parallel computing opens new horizons for both education and research."