Few dispute the importance of supercomputing to U.S. competitiveness. The argument is around whether current government efforts – primarily through the Advanced Scientific Computing Research (ASCR) program within the U.S. Department of Energy (DOE) – are effective and sufficient or wasteful and excessive.
Yesterday, a panel of HPC experts testifying at a U.S. House of Representatives hearing (Subcommittee on Energy – Supercomputing and American Technology Leadership) argued the decline in U.S. supercomputer research spending is putting U.S. computer and competitive leadership at risk.
“In the past century the federal government financially supported two-thirds of the nation’s research and development activity but that has gradually declined to one-third. Industry, on the other hand, has increased its share from one-third to about two-thirds. The problem is that, because of financial market pressure for rapid returns, industry focuses largely on ‘D,’ not ‘R,'” said panelist, Norman Augustine, retired chairman and chief executive officer of Lockheed Martin Corp.
“The result has been that in terms of arguably the most significant measure of national research investment, research funding as a fraction of GDP, the United States has recently dropped from first to seventh place in the world. The extent of America’s disinvestment in research is such that America now ranks 29th among developed nations in the fraction of research that is governmentally funded. It is projected that within about five years China will surpass the U.S. in both research funding as a fraction of GDP and absolute funding,” said Augustine.
DOE operates 17 laboratories located throughout the country, the efforts of which are principally focused on energy research and the provision of weapons that underpin the nation’s nuclear deterrent. FY2015 funding for ASCAR is $541 million.
Augustine contends that because DOE laboratories enjoy relatively stable funding and are well suited to “long-term, high-risk/high-payoff, often-large projects with applicability that may not be evident at their outset.” He cited support of research in commercial nuclear fusion and hydraulic fracturing to produce shale gas would be but two examples of such endeavors.
For the HPC community, these are familiar arguments and were shared by the other panelists:
- Dr. Roscoe Giles, Chairman, DOE Advanced Scientific Computing Advisory Committee, and Professor at the Department of Electrical and Computer Engineering.
- David Turek, Vice President, Technical Computing, IBM
- Dr. James Crowley, Executive Director, Society for Industrial and Applied Mathematics
Just getting to the hearing was challenging for one of the panelists (Dr. Giles) as recent snow in Boston restricted flights out. He participated by video conference.
The practical matter for the HPC community is prying loose government funding. Energy Subcommittee Chairman Randy Weber (R-Texas) issued a statement in seeming strong support of ASCR:
“As we face the reality of ongoing budget constraints in Washington, it is our job in Congress to ensure that taxpayer dollars are spent wisely, on innovative research that is in the national interest, and provides the best chance for broad impact and long-term success. The basic research conducted within the ASCR program clearly meets this requirement. High performance computing provides a platform for breakthroughs in all scientific research, and accelerates applications of scientific breakthroughs across our economy.”
At least as interesting as the discussion of funding was discussion around key technology challenges including the importance of co-design principles (simultaneously algorithm, software and hardware development) in supercomputing, worries over hitting the fundamental limits of silicon, and the difficulties faced in achieving exascale computing systems.
Technology transfer was another concern cited. Moving applications and technology out from national supercomputing centers into the mainstream can be challenging. Turek noted the general rule today is the national labs are 5-to-7 years ahead of the broader commercial HPC market. Panelists were pressed on their thoughts for how tech transfer could be accelerated while steering clear of problematic conflict of interest issues that complicate private-public collaborations.
On the whole, it was an interesting, if not entirely unfamiliar, conversation. Part of the purpose of the meeting, said Dr. Giles, was to renew interest in a bill passed by Congress but not the Senate last year. Subcommittee member Randy Hultgren (R-Ill.), had introduced a bill, H.R. 2495 (IH) – AMERICAN SUPER COMPUTING LEADERSHIP ACT, which eventually died in the Senate. Dr. Giles was an advisor on the bill.
Here are a few key points of the bill:
“…Amends the Department of Energy High-End Computing Revitalization Act of 2004 with respect to: (1) exascale computing (computing system performance at or near 10 to the 18th power floating point operations per second), and (2) a high-end computing system with performance substantially exceeding that of systems commonly available for advanced scientific and engineering applications.
Directs the Secretary of Energy (DOE) to:
(1) coordinate the development of high-end computing systems across DOE;
(2) partner with universities, National Laboratories, and industry to ensure the broadest possible application of the technology developed in the program to other challenges in science, engineering, medicine, and industry; and
(3) include among the multiple architectures researched, at DOE discretion, any computer technologies that show promise of substantial reductions in power requirements and substantial gains in parallelism of multicore processors, concurrency, memory and storage, bandwidth, and reliability.” Click for a Library of Congress summary of the bill.
A video archive of the session is available as are copies of panelists’ written statements. Members of the subcommittee include: Randy Weber (R-Texas), Chair; Dana Rohrabacher (R-Calif.); Randy Neugebauer (R-Texas); Mo Brooks (R-Ala.); Randy Hultgren (R-Ill.).