Skip to main content

Distributing computing resources: The social challenge

May 10, 2005

By bringing together a diverse group of scientific interests to share a large, distributed computing resource, the Grid Laboratory of Wisconsin (GLOW) itself has become an important subject of research.

“It becomes a subject of study,” says Miron Livny, the computer science professor who has helped bring together an assortment of research interests under the GLOW banner. “For my team (in computer science), GLOW is a laboratory to figure out how our technology is working. Having an engaged community of users is of tremendous value to my research.”

But perhaps more importantly, asserts Livny, the GLOW project has become a way to put interdisciplinary interaction in the sciences under the microscope of the sociologist. UW–Madison sociologists Daniel Kleinman and Nicole Kaufman have become integrated into the GLOW consortium and are assessing the social aspects of divergent research groups sharing a single computing resource.

“This work gives us a unique perspective,” says Livny. “We’re not just running with the technology.”

Perhaps the primary social challenge of a project like GLOW, says Kaufman, is that it revolves around a finite resource: available processing power. “You have relative resources and one of the biggest problems is in assigning priority. All (of the groups) have really heavy demands and there is a lot of pressure to do the research.”

Indeed, some groups are more engaged than others, and figuring out ways to parse available computing cycles can be a critical management issue on any given day. How much some groups participate in GLOW, Kaufman asserts, becomes a determinant of the advantages GLOW might confer.

But that very problem, asserts Livny, is why the work of Kleinman, a professor of rural sociology, and Kaufman, a graduate student in sociology, becomes essential.

“We can take care of the technology, but building a partnership is not a technical question.” Livny says. “I find the sociology to be extremely important for the success of this kind of technology. Why does it make sense for all of us to work together?”

One big reason, suggests Kaufman, is that even though the GLOW resource is finite and users must allot and share time, the advantages are alluring.

It may not be the case, however, that all groups benefit equally, as some tend to be heavier users and others are not yet fully conversant with the technology. The groups also may not have the same volume of computing needs.

“It’s striking that the highest users have great data processing demands, and often have been using Condor (the computer template devised by Livny to scavenge idle processing power) already; not everyone has used the GLOW resources as aggressively,” Kaufman says.

“It is relevant, though, that this sort of resource sharing is happening at all. Although some scientific fields have a long tradition of sharing expensive equipment, computing has arguably become the newest face of shared resources.”