By Salvatore Salamone
Feb. 27, 2008 | For years, technology developed to meet the computing requirements of Sandia, Oak Ridge, and other national laboratories has pushed the high-performance computing (HPC) envelope and spurred innovation in a broad range of scientific research areas. In fact, over the years much of the technology developed for the labs has been rapidly incorporated into commercial systems used to advance life science research.
Keeping that tradition alive, Sandia and Oak Ridge this month launched the new Institute for Advanced Architectures, a joint initiative for the development of a next-generation HPC system dubbed an exascale computer.
“An exascale computer is essential to perform more accurate simulations that, in turn, support solutions for emerging science and engineering challenges in national defense, energy assurance, advanced materials, climate, and medicine,” said James Peery, director of the Computation, Computers, Information, and Mathematics Center at Sandia.
As the name exascale implies, the systems will operate in the exaflop range, where an exaflop is a quintillion—10 to the 18th power—operations per second. To put this number in perspective, an exaflop system would deliver one thousand times more operations per second than a petaflop system, which was the objective of IBM’s Blue Gene project when it was launched about eight years ago. (A petaflop is a quadrillion—10 to the 15th power—floating point operations a second.)
An exascale system would bring significantly more processing power to scientific research than today’s computers. For example, the most powerful computer in the world today is the IBM eServer Blue Gene Solution at the Lawrence Livermore National Laboratory, which has a peak benchmarked performance of 478.2 teraflops. (A teraflop is a trillion—10 to the 12th power—operations per second.)
Adding More than Muscle
One goal of the institute is to “to close critical gaps between theoretical peak performance and actual performance on current supercomputers,” said Sandia project lead Sudip Dosanjh. “We believe this can be done by developing novel and innovative computer architectures.”
For instance, one area the institute will focus on is reducing or eliminating the growing mismatch between data movement and processing speeds. “In an exascale computer, data might be tens of thousands of processors away from the processor that wants it,” said Sandia computer architect Doug Doerfler. “But until that processor gets its data, it has nothing useful to do. One key to scalability is to make sure all processors have something to work on at all times.”
As has been the case with each successive HPC generation, the interplay between hardware and software will be taken into account. “In order to continue to make progress in running scientific applications at these [very large] scales, we need to address our ability to maintain the balance between the hardware and the software.” said Jeff Nichols, who heads the Oak Ridge branch of the institute, “There are huge software and programming challenges, and our goal is to do the critical R&D to close some of the gaps.”
Additionally, the institute plans to address the electrical power issue, which is a growing problem in many HPC environments. If current technology and approaches were used, an exascale computer would need tens of megawatts of power and cost millions of dollars for electricity to power and cool the system, according to the labs. “We want to bring that down,” said Dosanjh.
Congress has set aside $7.4 million in funding for the Institute for Advanced Architectures in fiscal year 2008.
What HPC technologies are you most interested in today? Are you using systems with quad-core processors, FPGAs, or other accelerators? Is power consumption and cooling an issue? What technologies on the horizon do you think will be the most helpful in meeting your future computing requirements? Drop me a note at firstname.lastname@example.org and share your thoughts on the subject.
This story first appeared in Inside IT,one of Bio-IT World’s free e-newsletters. Subscribe here.