For over half a century, America has led the world in high performance computing (HPC) thanks to sustained federal government investments in research and the development and regular deployment of new systems, as well as strong partnerships with U.S. computing vendors and researchers.
Within the federal government, the Department of Energy (DOE) leads this effort, with the nation’s fastest and most capable supercomputers housed at the DOE national laboratories.
As of the November 2018 TOP500 list of the world’s most powerful supercomputers, the U.S. owns five of the world’s top ten systems. Summit at Oak Ridge National Laboratory (ORNL) ranks as number one, Sierra at Lawrence Livermore National Laboratory at two, Trinity at Los Alamos at six, Titan at ORNL at nine, and Sequoia at Lawrence Livermore National Laboratory at ten. It is not just hardware that defines American leadership. U.S. leadership also depends critically on major ongoing efforts in key fields of research, including applied mathematics and computer science, advanced lithography, and nanoscale materials science—areas where DOE Office of Science research programs excel.
The strong synergy between hardware development, on the one hand, and software and application development, on the other, has been a defining strength of the U.S. approach—and has transformed HPC over the years into an ever more capable tool of both science and industry.
America has reaped enormous benefits from this leadership position. HPC has been a major factor in accelerating U.S. scientific and engineering progress. It has enabled the development of advanced manufacturing techniques and rapid prototyping. And it has ensured national security through stewardship of the nation’s nuclear stockpile in the absence of testing.
America has so dominated the HPC scene that it is sometimes easy to take for granted the enormous economic and competitive benefits that flow from this global leadership position.
Today, however, America’s leadership in HPC is under challenge as never before, as nations in Asia and Europe invest heavily in HPC research and deploy new systems.
The top priority today is the continued progress to exascale—the next level of supercomputing, defined by processing at a quadrillion, or billion billion, calculations per second.
The U.S. is on track to deploy its first exascale system at Argonne National Laboratory in 2021 and a second exascale system at ORNL expected in the 2021-2022 timeframe. But equally important will be the ability to harness this system for science and industry through software and application development. To this end, the Office of Advanced Scientific Computing Research with the DOE Office of Science is partnering with the National Nuclear Security Administration within DOE in supporting the Exascale Computing Project (ECP). Supporting researchers across the nation, the ECP is sponsoring the integrated application, hardware, and software research and development necessary to effectively use an exascale system.