Houghton--The computational science and engineering PhD program has received a three-year, $758,000 grant from the NASA Goddard Space Flight Center to support the appointment of parallel computing expert Phillip Merkey as a research assistant professor in the computer science department.
Merkey is widely known for his work on the development of Beowulf cluster computing platforms. A Beowulf cluster is a collection of a few tens to a few hundreds of PCs connected to each other by off-the-shelf networking equipment to form a parallel computer with just one keyboard, one monitor, and one network connection. Beowulf clusters are used to solve very complex scientific and engineering problems but are much less expensive than parallel computers built of custom hardware and software. Later this year, Merkey will bring a Beowulf cluster to the Department of Computer Science for use in his research, as well as to support other computer science projects and the computational science and engineering PhD program.
"It is exciting to have Phil join the computer science faculty," Department Chair Linda Ott said. "We have been involved in high-performance and parallel computing since the mid-1980s. The new Beowulf cluster and Phil's expertise in science and engineering applications will be a great addition to research efforts in this area."
Another example of computers working in tandem is the SETI@home project, in which PCs worldwide are loaded with special software to analyze data from the Search for Extraterrestrial Intelligence. Their differences are significant, however. SETI is considered a speculative long-shot while Beowulf clusters work on real-world problems. But they have a peripheral similarity. "We are in the same ballgame," said Associate Professor Steven Seidel, director of the computational science and engineering PhD program and co-principal investigator on the Beowulf project. "You have lots of computers and you try to get them to work on the same problem and cooperate with one another."
While at MTU, Merkey continues the work he began with the NASA Earth and space sciences program, testing new parallel programs that make it possible for the individual computers in Beowulf clusters to work together. Merkey will also visit other Beowulf sites across the US, helping principal investigators evaluate their own experimental programs.
Ultimately, Merkey's research should help NASA reach its objective of designing a parallel computer so powerful that, by the year 2007, it can solve a quadrillion (that's one million billion, a 1 followed by 15 zeros) arithmetic operations per second. The overall goal is to predict how such computer architectures will impact the scientific community.
Seidel expects the Beowulf project to yield multiple benefits. "Phil can do a lot to help establish the CS&E program as the center of high performance computing at MTU," he said. "Plus, this project will attract high-quality graduate students and position the CS&E program to contribute to additional externally funded, computationally intensive research projects."
Although Merkey will be on the road much of the time at the beginning of this project, it will not be because he is unaccustomed to UP winters. He received his BS in Mathematics from MTU in 1977. Since then, he earned a PhD in Mathematics from CalTech, worked for the Institute for Defense Analyses, and taught at the University of Maryland's Baltimore County campus.
For more information on Beowulf cluster technology, visit http://www.beowulf.org. For information about the computational science and engineering PhD program, see http://www.cs.mtu.edu/grad/CSE.