Gordon Bell Prize for High Performance Computing

PITTSBURGH: A team of researchers from Carnegie Mellon University, the University of Pennsylvania and the Pittsburgh Supercomputing Center (PSC) won the 2003 Gordon Bell Prize, one of the most prestigious awards for high performance computing.

The team was honored for developing earthquake computer simulations that play an important role in reducing seismic risk. Team members included Volkan Akcelik, Jacobo Bielak, Ioannis Epanomeritakis, Antonio Fernandez, Omar Ghattas, Eui Joong Kim, Julio Lopez, David O' Hallaron and Tiankai Tu of Carnegie Mellon; George Biros of the University of Pennsylvania, and John Urbanic of the Pittsburgh Supercomputing Center.

"This is a great honor for a team that has worked to accomplish major advances in our ability to model and understand earthquake behavior," said Chris Hendrickson, head of Carnegie Mellon's Civil and Environmental Engineering Department. "Over a period of 10 years, they have collaborated on a series of increasingly ambitious and influential computer models of earthquake behavior, creating fully realistic three-dimensional representations of complex basin geology, earthquake sources and earthquake ground motion."

John L. Anderson, dean of Carnegie Mellon's College of Engineering, said the award is another example of the universitys successful interdisciplinary problem-solving environment. "The project draws upon expertise in computational science and engineering, computer science, earthquake engineering and seismology," he added.

The Quake Project's large-scale models and computer simulations have pushed the capability of existing hardware and software systems. "The Bell Prize recognized our recent Los Angeles Basin earthquake simulations on PSC's 3000-processor LeMieux supercomputer," said Jacobo Bielak, professor of civil and environmental engineering at Carnegie Mellon. "These simulations provide unprecedented levels of resolution and detail, and were enabled by multi-resolution wave propagation methods we have developed. Conventional techniques would have required 1,000 times more computing power to achieve the same accuracy," Bielak said.

One of the keys to making such large-scale simulations possible is the ability to create extremely large models of the Los Angeles Basin. "We have developed special algorithms and data structures that have allowed us to generate models containing several billion variables," said David O'Hallaron, associate professor of computer science and electrical and computer engineering at Carnegie Mellon. "These are among the largest models that have been generated in any field."

In addition to modeling earthquakes, the Bell Prize recognized the group's work on methods for determining subsurface geology from observations of surface ground motion due to past earthquakesthe so-called inverse problem.

"We've been able to solve inverse problems on an order of magnitude larger and more complex than any previously attempted," said Omar Ghattas, professor of biomedical engineering and civil and environmental engineering at Carnegie Mellon. "We had to develop new inversion methods that could scale to the millions of parameters characterizing such problems." "At PSC, we're gratified that the Quake Group has received this recognition," said Pittsburgh Supercomputing Center scientific directors Michael Levine and Ralph Roskies in a joint statement. "This research has important social impact. In short, it will save lives. It demonstrates the scientific contribution of high-end computational systems, such as LeMieux, and the value of close collaboration among computer scientists and domain scientists." John Urbanic, a PSC staff computational science consultant, said it wouldn't have been possible without a system like LeMieux.

The Gordon Bell Prize, given each year at the annual Supercomputing Conference, was established in 1988 by Gordon Bell, a pioneer in computer architecture who taught engineering and computer science at Carnegie Mellon from 1966 to 1972. Bell, who spent 23 years at Digital Equipment Corp. as vice president of research and development, is a senior researcher in Microsoft's Media Research Group, part of the San Francisco Bay Area Research Center, which maintains an interest in startup ventures. The Gordon Bell Prize winners were announced Nov. 20 at the 2003 Supercomputing Conference in Phoenix, Ariz.

More information on this group's work can be found at www.psc.edu/science/2003/earthquake/big_city_shakedown.html.