High-Performance Computing | News
U Tennessee Prof Takes on Exascale Computing
- By Dian Schaffhauser
A computer science professor from the University of Tennessee in Knoxville has just earned a million dollar grant to explore the next generation of high performance computing. Jack Dongarra, who's also affiliated with the University of Manchester and Oak Ridge National Laboratory, received the three-year grant from the United States Department of Energy to better understand the changes that will be required in the software for exascale supercomputers, as they're called.
This new generation of supercomputers will be capable of a quintillion floating point operations per second or one exaflop — equivalent to a thousand petaflops. Today's fastest supercomputer, China's Tianhe-2, performs at about 34 petaflops per second; Cray's Titan supercomputer was benchmarked at 18 petaflops per second. Exascale computing is expected to be reached much later in this decade.
However, said Dongarra in a statement, "You can't wait for the exascale computers to be delivered and then start thinking about the software and algorithms." The challenges for what he calls "extreme computing" include programming issues, fault tolerance, and power usage.
Dongarra, who continues to teach and contributed to 46 peer-reviewed papers in 2012, is working with researchers on several problems. One project is called the Parallel Runtime Scheduling and Execution Control (PaRSEC). PaRSEC is a generic framework with libraries, a runtime system, and development tools to help developers port their applications to new kinds of environments.
Dongarra is also developing an algorithm to overcome a reliability problem associated with the increasing number of processors. Now, when one processor fails, the calculation may have to be repeated in part or fully. The algorithm project aims to develop software that can survive failures and perform auto-tuning to adapt to the hardware.
"The exascale computers are going to be dramatically different than the computers we have today," he noted. "We have to have the techniques and software to effectively use these machines on the most challenging science problems in the near future."
Dian Schaffhauser is a writer who covers technology and business for a number of publications. Contact her at email@example.com.