The House of Representatives passed an exascale computing support bill Monday highlighting once again the link between next-generation HPC and national competitiveness. The American Super Computing Leadership Act of 2014 (H.R. 2495) seeks to “amend the Department of Energy High-End Computing Revitalization Act of 2004 to improve the high-end computing research and development program of the Department of Energy, and for other purposes.” Like other technology-focused initiatives that promote American leadership, the bill seems to have some bipartisan appeal. Sponsored by Illinois Republican Randy Hultgren, 15 of the bill’s 22 cosponsors are Democrats.
Despite exascale computing being cited as a critical factor to American competitiveness by almost everyone with a stake in the game and despite the serious obstacles that must be overcome to field these next-generation machines, the political will to institute and fund the necessary R&D has been lacking. Notably the American Super Computing Leadership Act does not authorize appropriations, but it does put forth a plan for the development of exascale technologies and systems capable of advancing the mission of the DOE.
Rep. Randy Hultgren (R-Ill.), who proposed the bill, cited some of the key factors behind the push toward more powerful computing systems. “Unfortunately, America is falling behind as China boasts the world’s fastest computer and is not slowing down,” he said. “Massive gains in computing power are necessary to meet our national security, scientific, and health care needs. Testing our nuclear stockpile as it ages will require more and more advanced computer simulations. Advanced modeling of the human brain gives scientists a better understanding of what causes Alzheimer’s and cardiovascular disease, and what can treat it. Faster computing power speeds up drug development to get cures for our nation’s top killers to the marketplace faster so Americans receive better health outcomes.”
Introduced on June 25, 2013, the bill would amend portions of the 2004 law to be more in line with the current challenges of extreme-scale computing. In 2004, the community was on the brink of the petascale barrier the breaking of which (in 2008) turned out to be more evolution than revolution. But getting to exascale is no longer a matter of scale. There are serious power/memory/programming/ resiliency constraints and the cost model has changed from one of expensive FLOPS and nearly free data movement to cheap FLOPS and expensive data movement.
If passed by the Senate and signed by the president, H.R. 2495 would direct the Secretary of Energy (DOE) to do the following:
- Carry out a coordinated research program across the Department to develop exascale computing systems.
- 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.
- 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.
- Establish two or more National Lab-industry-university partnerships to conduct integrated research, development, and engineering of multiple prototype exascale architectures.
Upon the bill’s enactment, the Secretary will have 90 days to submit an integrated strategy and program management plan that includes target dates as well as an estimate of costs. Regular status reports will be due at the time of budget submission each fiscal year.
The DOE will be tasked with conducting mission-related co-design activities in developing these exascale platforms. Co-design is defined by the bill’s authors as “the joint development of application algorithms, models, and codes with computer technology architectures and operating systems to maximize effective use of high-end computing systems.”
Although the 2004 law makes reference to industrial partnerships, the new act clarifies the need for outreach to “domestic industries, including manufacturers” – a key addition for the bill’s sponsor Rep. Randy Hultgren (R-Ill.).
“This legislation would ensure better access for the research community and industry,” said Hultgren. “Thus, our manufacturers in Illinois who employ thousands of my constituents can improve the prototypes they put off their lines, producing a better product and more opportunities to expand their businesses and hire more workers.”
This is not the first time the 2004 law has been under review. On July 31, 2012, Sens. Jeff Bingaman (D-NM), Lamar Alexander (R-TN) and Dick Durbin (D-IL) introduced the Department of Energy High-End Computing Improvement Act of 2012 (S-3459), seeking to update the High‐End Computing Revitalization Act of 2004 to preserve DOE HPC and to establish the exascale initiative.
Unlike the current proposed update, that bill would have authorized the appropriations needed to carry out the act (for fiscal years 2013 through 2015), allocating $110 million the first year, $220 million the second year and $300 million the third year.
For reference, the 2004 act authorized $50,000,000 for fiscal year 2005, $55,000,000 for fiscal year 2006 and $60,000,000 for fiscal year 2007.