UT Professor Leads World Effort in Developing Next Generation of Supercomputers
KNOXVILLE — East Tennessee is home to Jaguar and Kraken, two of the fastest supercomputers in the world. These high-performance systems help solve the world’s most challenging problems by running simulations of what might happen in the real world.
Jack Dongarra, distinguished professor of electrical engineering and computer science at the University of Tennessee, Knoxville, knows the power of these computers well. He puts together the much anticipated biannual top500 list, which ranks the 500 fastest supercomputers in the world.
“Supercomputers are critical for national security issues. They make fundamental contributions to the economy and the competitiveness in engineering and manufacturing,” Dongarra said.
Speed is integral to making supercomputers “super.” Right now, their speed is measured in petaflop per seconds or petaflop/s. A petaflop is 10 to the 15 calculations per second. Jaguar is currently the second-fastest computer in the world, operating at 2 petaflop/s.
To put that in perspective, you would need three Neyland stadiums, each full of a 100,000 people with interconnected laptops working simultaneously on a single computation to equal the power of Jaguar.
But that power’s not enough.
Today, full simulations can’t be completed for many types of problems. Science needs more power, more speed and more memory, or else scientific advancements will taper off.
Supercomputers need to be taken to the next level. This level is called exascale.
“The drive to exascale is coming about from the science community. It’s not the technology that’s drawing us to exascale. The technology can take us there and that’s the good news, but it’s really the science that’s the driver, in some sense. These science applications have stepped up and said, ‘In order for us to do the kinds of problems that we can’t do today — in order for us to do them in the future — we need exascale computing,” Dongarra explained.
Exascale is three orders of magnitude higher than a petaflop. It is 10 to the 18 calculations per second. So that would be 3,000 Neyland Stadiums full of 100,000 people working simultaneously on interconnected laptops on a single computation.
The task is so important and so challenging, computing leaders who normally would be competing to build the fastest supercomputer are coming together to draw a roadmap for the next generation of supercomputers.
Leading this international effort is Dongarra.
“You can’t wait for the exascale machine to be delivered and then start thinking about the software and algorithms. The exascale systems are going to be dramatically different than the systems we have today. We have to have the techniques and software to effectively use these machines on the challenging science problems of the day,” Dongarra said.
Dongarra says exascale computing is crucial to science advancements, national security and the economy.
“It’s also critical with respect to the economy and the manufacturing of various goods and technology in the U.S. If you think about companies that have these high performance systems, they are using them to develop better products,” he said.
Achieving exascale is hampered by cost and technology, but Dongarra circles the globe several times a year, meeting with the brightest computing minds around. They feel confident they will achieve 10 to the 18 by 2020 — and the world will be better for it.