When the world’s most powerful supercomputer goes online in 2011, it won’t come pre-installed with user-friendly software applications. Not to worry! To solve that problem, The Great Lakes Consortium for Petascale Computation will spend the next three years ramping up for the world’s first sustained petascale computational system by developing new computing software, applications and technologies designed for open scientific research.
The Great Lakes Consortium is the result of collaboration among colleges, universities, national research laboratories and other educational institutions dedicated to the Blue Waters Project.
The Blue Waters Project, based at the University of Illinois at Urbana-Champaign’s National Center for Supercomputing Applications, will build a machine in conjunction with IBM capable of sustaining computations of one to two petaflops – computing parlance for 1 quadrillion calculations per second – on many practical scientific and engineering applications.
The consortium’s ultimate goal is for Blue Waters to be fully user-friendly for scientists across the country, so when it launches, it will include intense support for application development, system software development, interactions with business and industry and educational programs.
Iowa State University researchers Srinivas Aluru, Mark Gordon and James Oliver say they’re eager to help the scientific community step into what they call the second revolution in information technology.
Aluru, a Stanley Chair in Interdisciplinary Engineering and a professor of electrical and computer engineering, will direct ISU’s work with the consortium.
“The dramatic increase in computing capability makes this project a national asset,” he said. “A lot of money will be poured into this research. To justify public expenditure we want to be ready.”
The National Science Foundation is supporting the supercomputer project with a $208 million grant, said Aluru, whose research group has used supercomputing power to help with the recently concluded effort to sequence the corn genome. To do that, they developed software that uses thousands of processors to build genome assemblies in days instead of months.
And now Aluru is ready to make the leap to even more powerful computing. But before that can happen, researchers must work out the bugs and bottlenecks that petascale computational levels might present.
The issue is not just Blue Waters’ peak potential, but its sustained capacity while solving problems, he said.
“That efficiency depends on the code we write,” he said. “We need to find the way to get higher than 70 percent efficiency on solving several challenges.”
Mark Gordon, ISU’s Frances M. Craig Distinguished Professor of chemistry and the director of the applied mathematics program for the U.S. Department of Energy’s Ames Laboratory, said parallel computing in chemistry, for example, has used, at most, clusters of 32-128 computers for supercomputing challenges for the past 15 to 20 years. Researchers therefore haven’t had the hands-on opportunity to work through the potential bottlenecks for using up to 100,000 clusters.
“It’s a whole new ballgame with new bottlenecks,” he said. “When you move toward the petascale range, we might run up against physical limitations, such as the speed of light. And the communications and data sharing issues increase by orders of magnitude. We’ll need an efficient way of communication and comparing and collecting.”
One of the consortium’s strategies will be forming petascale application collaboration teams or PACTS, Aluru said.
“Each team will work on individual problem to figure out how to use the petascale computer and avoid mistakes,” Aluru said.
Aluru said the Nation Science Foundation-funded project will provide two “step-up machines” along the way.
James Oliver, the director of ISU’s CyberInnovation Institute, said the jump to petascale computing power calls for tools such as C6, ISU’s six-sided virtual reality room that displays computer-generated images at the world’s highest resolution. He said C6 would be an ideal place to build interfaces that can display and work with all the data produced by the supercomputer.
Aluru said the consortium held its inaugural meeting this week to begin to lay out the technical challenges it faces. Back at ISU, Gordon said he’s waiting for word from the National Science Foundation to grant his team early access to the Blue Waters team and hardware.
“We’re looking forward to trying out our ideas to see if they’re going to work.”