UMaine Computer Engineers Helped Lay Foundation For New U.S. Army Supercomputer

University of Maine computer engineers and employees of Applied Thermal Sciences of Sanford, Maine have bragging rights in a recent announcement that the U.S. Army will create what may be the second most powerful computer in the world. The Army decision to build the machine known as the MACH 5 (Multiple Advanced Computers for Hypersonic, G5) is based in part on UMaine research funded by the Army to develop more powerful and less expensive computing technology for military research. In May, the Army announced that it would purchase a system consisting of 1,566 Apple Xserve G5 desktop computers for $5.8 million. When Army contractor, Colsa, Inc. of Huntsville, Alabama, completes the system this fall, it is expected to be second in speed only to the $350 million Earth Simulator computer in Japan. The Colsa computer is expected to operate 25,000 times faster than the average desktop computer. “The goal of our work is to learn how to build bigger, better, faster, and cheaper supercomputer technology for the Army,” says Eric Wages, supercomputer engineer in UMaine’s Dept. of Industrial Cooperation. Working with Wages at the Target Technology Center in Orono are Bruce Segee and Andy Sheaff of UMaine’s Dept. of Electrical and Computer Engineering; Glen Beane, a graduate student in the Department of Computer Science; and Larry Thompson of Applied Thermal Sciences, an engineering research firm in Sanford. For the last two years, this team has been working to analyze the performance of complex software that simulates high-speed missile trajectories on a cluster of high performance desktop computers. The software is so complex that engineers have taken what amounts to a digital magnifying glass to see how various sections of the code perform. “Our job is to profile the code and then figure out how to make it run faster without changing it,” says Wages. Scientists typically use supercomputers to understand complex phenomena such as nuclear explosions, global climate or the Big Bang. A faster computer essentially allows scientists to work on complicated problems with greater accuracy. At the Target Technology Center in Orono, the UMaine team is currently using a supercomputer cluster of 208 desktop computers, or nodes. Starting this week, researchers will begin building a new 256 node Apple Xserve G5 system, “a baby MACH 5,” says Wages. The new $680,000 system is funded by the Army under contract with UMaine. The current supercomputer system will remain at Target. “We hope to keep the system running 24/7 and available for UMaine researchers and businesses,” says Wages. UMaine students and faculty in chemistry, marine sciences, computer science and computer engineering are already using the system for their research. For more information on the UMaine's supercomputers or the people behind the project, see http://www.clusters.umaine.edu.