Features:

HIGH PERFORMANCE COMPUTING CRITICAL TO NATIONAL DEVELOPMENT
by Tim Curns, Editor

The constant battle between supercomputer vendors and research institutes to create the fastest supercomputers in the world is not a new one, but it is definitely heating up. With the NASA/ SGI/Intel Columbia project, in addition to IBM's BlueGene/L, the U.S. boasts two of the fastest supercomputers in the world.

But how important is our HPC competitiveness to industry and research?

At SC2004, the Council on Competitiveness announced that it will sponsor a panel on the importance of advanced computing to U.S. competitiveness entitled, "Leadership in a Global Economy: 'Out-compete' means 'Out-compute.'" The panel will focus on the linkages between industrial use of this critical technology and competing successfully in a global economy. The panel is on Tuesday 11/9 at 3:30pm in room in room 317-318.

HPCwire editor Tim Curns had a chance to speak with Deborah Wince-Smith, President of the Council on Competitiveness, Suzy Tichenor, Vice President and Director for the High Productivity Computing Project, and David Shaw, chairman of the panel at Supercomputing 2004 and chairman of D.E. Shaw & Co., about the role of competitiveness in U.S. high-performance computing initiatives.

The Council on Competitiveness believes that high-performance computing is one of the "transformational tools" needed to drive innovation, productivity and competitiveness. The panel will be an assemblage of various industry HPC users, who will discuss some of the Grand Challenge tasks that require increasing amounts of compute capability. The group will also speak to the economic ramifications of solving these tasks.

"In recent years, interest has revived in attacking some of these more challenging problems using novel supercomputer architectures, driven in part by global economic competition, new challenges to national security, and the rising technological capabilities of other nations. The technical advances that will be reported at SC 2004 assume special significance in the context of these global challenges," said David Shaw.

According to the Council, in order for the U.S. to be able to compete successfully in global markets now and in the future, the nation must transform into an innovation based economy.

But how does this happen?

Suzy Tichenor said that the U.S. must think about what kind of infrastructure is needed, and what kind of resources are needed to support that. "We feel very strongly that high-performance computing is one of those key resources that is going to have to be widely accessible in order to drive innovation, and therefore, productivity, growth and ultimately, our standard of living," she said.

A Council survey this summer found that HPC users have Grand Challenge problems that cannot be solved with the systems in place today, signifying a need for more powerful, and more importantly, more productive computational systems. The Council, not made up of computer scientists to judge the superiority of one architecture over another, seeks instead to identify the way scientists should use these tools as drivers for innovation and competitiveness.

What role does the government have in all this? Some have lost faith in recent government initiatives like the High-End Computing Revitalization Task Force (HECRTF), citing lofty ideals with little funding actually becoming a reality. And while money for revitalizing high-end computing may not be ready yet, the Council on Competitiveness believes having the initiative itself is a step in the right direction.

"We are working closely with Dave Nelson, over at the NCO/IT R&D (National Coordination Office for Information Technology Research and Development), the Department of Energy, Office of Science, and DARPA," said Tichenor. "What we hope we can contribute is a better understanding of what the private sectors' needs are, and what the drivers are within the private sector for using HPC. Also, we hope to stimulate greater partnership and collaboration between the public and the private sector on the development and application of HPC."

Deborah Wince-Smith added that "our prosperity and security depend on aggressive use of high performance computing resources. These are transformational tools and it's exciting to see agencies such as DARPA and DOE investing in this technology and reaching out to partner with the private sector to make available resources and expertise that might not be available otherwise."

The Council suggests that the government, as a leading customer and user of HPC, helps to drive the innovation process. Initiatives like the National Leadership Computing Facility at Oak Ridge National Laboratory illustrate the government's growing investment in the technology of tomorrow. The government thus creates a virtual symbiotic relationship with industry, producing a double return on the investment.

As far as government is concerned, the Council predicts that the recent elections and subsequent Republican control of Congress should not affect federal HPC initiatives at all. The Council, a non-partisan, non-lobbying organization, has found that both Democrats and Republicans recognize and understand the need for investment in innovation tools.

There are barriers, however, to ensuring U.S. HPC competitiveness. The lack of computational scientists, for instance, is a growing problem. The Council hopes to push collaboration with the educational system to ensure adequate training in computational science, rather than mere computer science.

Another obstacle involves company perspective. Does a company see high- performance computing as a cost or an investment? An investment perspective, according to the Council, could be a major driver in the types of systems available to their researchers.

"Leaders in both the public and private sectors have proposed exploring the extent to which both public- and private-sector requirements might be addressed by a common class of novel architectures capable of performance far beyond the capabilities of current high-end computing facilities," said Shaw.

The Council on Competitiveness hopes the SC2004 panel will examine some of these barriers and discuss the ways in which industry, academia, and government can partner to propel the country ahead in a competitive environment.