Institute Provides
Intensive Computing

Computational scientists focus on solving real science and engineering problems. These problems have become very complex, creating the need for the Computational Science and Engineering Research Institute. Steve Seidel (left) and Phil Merkey lead the institute. The Cray supercomputer is behind them.

by Dean Woodbeck

Imagine 64 computers, all tied together, diligently working on the same massive computational problem. Each has its own part of the job to do, working simultaneously with the rest.

Such an arrangement allows for more complex problems completed much more quickly.

Michigan Tech’s array of computers is called a Beowulf cluster and is one of the campus-wide resources made available by the Computational Science and Engineering Research Institute.

The Institute has also acquired a Cray T3E supercomputer to support high-powered computing needs.

Computational scientists focus on solving real problems in science and engineering.

“Take global warming,” says Steve Seidel, professor of computer sciences and one of the computational science institute’s creators. “All of the predictions are done based on computer models. But you don’t just go out and buy a global warming software program—someone has to create the software. That’s a computational scientist.

“The genome project was a huge computational project. Computational scientists had to create that sophisticated program to allow the sequencing research to happen.”

The Computational Science and Engineering Research Institute provides a home for interdepartmental science and engineering research that requires powerful computers.

Some of the research using these resources included modeling processes in the atmosphere, environmental protection problems and a new language for programming computer networks like the Beowulf cluster.

“It doesn’t make sense to have such machines spread all over campus,” Seidel said. “It makes sense to have these machines together in a central location, but available to all.“

Phillip Merkey, a research assistant professor in the Department of Computer Science, is the institute’s first director. He came to Michigan Tech from the NASA Goddard Space Center and brought the Beowulf cluster with him.

Left: Graduate students Xiaodi (Lisa) Li and Jeevan Savant work in the institute’s nerve center, home of the Beowulf cluster and a Cray supercomputer.

“NASA made the Beowulf cluster available under a government program that allows for the redistribution of such equipment,” Merkey said. “NASA wasn’t using this any longer and we had a need at Tech.”

The research institute also supports a non-departmental PhD program in computational science and engineering. Under the program, a physics student, for example, continues to reside in the physics department, but does course work and research activities in the institute.

The PhD student becomes a computational scientist with an expertise in another area, like physics or chemistry or environmental science. The program currently enrolls students from geological engineering, computer science, and math.

Some of the projects conducted through the institute include:

	Adrian Sandu, assistant professor of computer science is developing better models for tracking atmospheric pollution.
	Seidel and Merkey are developing a portable version of the Unified Parallel C (UPC) programming language that allows computers in clusters (like Beowulf) to talk to each other.
	Foresters are pursuing projects in remote sensing.
	Geologists are modeling how contaminants spread below-ground and are developing techniques to improve groundwater contamination clean-up strategies.
	Mechanical engineers use the institute’s resources for their studies of computational fluid dynamics. Donna Michalek, assistant professor of mechanical engineering, is creating and simulating CFD tools for use in fuel injector design.

For more information on the Computational Science and Engineering Research Institute, see http://www.cse.mtu.edu/ or contact Phillip Merkey (906-487-2209 or merk@mtu.edu).