A three-year, $530,000 grant from the National Science Foundation will give Binghamton University researchers better access to technology that
will help in solving computationally demanding problems. The grant, combined with approximately $100,000 from the Binghamton University Research Foundation, will be used to purchase a 64-node high-performance cluster – basically a group of computer processors that can work together — for the Computer Science Department’s Grid Computing Research Laboratory.

“We’ll probably be able to double the capacity we have,” said Michael Lewis, an associate professor and one of five principal investigators on the grant.
The equipment will likely be ready for use starting in the spring semester. Lewis and his group are eager to put the new cluster to the test and hope to find new partners on campus.

Researchers who could use this technology include geologists simulating wave propagation to predict areas of high shake damage and computational chemists modeling the properties of marine enzymes.

“We built in this idea that this will enable and promote collaboration at Binghamton and at other institutions,” Lewis said. “We hope that we have the expertise and interest already here to work with researchers. What’s coming is an even larger-scale facility that will be able to
solve even larger problems.” Cognitive scientist Kenneth Kurtz, an assistant professor in the
Psychology Department, has already begun working with students and faculty members in the Computer Science Department, including Lewis.

Kurtz’s research involves the use of computational modeling to simulate cognitive processes. He wants to find out how people learn, how they make inferences and how they understand everyday experience. He’s never worked with grid computing before.

Kurtz builds neural networks – a form of computational system motivated by how the brain works. The runtimes can be long and the networks need to be tested repeatedly under different conditions, and grid computing offers an ideal way to do that.

“It’s a powerful time-saving technique for my research,” he said. Another aspect of his work wouldn’t be happening if it weren’t for his collaboration with the computer scientists, he said. “We’re starting to develop a machine learning system that integrates the modeling approach
I use for cognitive simulation and some of the parallelization techniques that they use,” Kurtz said.

Lewis’ role in this collaboration and others like it will be to create “middleware,” essentially software that allows a researcher to interact with the computers.

Before receiving this grant, the Computer Science Department already had several smaller clusters; a few other labs on campus run their own. Some are elegant stacks of computer processors contained in cabinets the size of a refrigerator. Another, less sophisticated, cluster is basically eight desktop computers wired together. Lewis notes that the field is young, perhaps 10 years old, and that early on it was an academic exercise that counted on small computers getting more and more powerful. That has come to pass, but researchers are still figuring out ways to harness this power for practical problem
solving.

Mainframe computers long were the dominant option for researchers, Lewis
said. But today some people believe grid computing is more promising because of the sheer volume of resources that can be contained in a grid. “The hope,” he said, “is that you can solve bigger, harder and more
interesting problems.”

No related articles were found