Need for supercomputer centers being reviewed

By Juan Antonio Osuna
CRA Staff

The National Science Foundation formed a task force in January to consider the future of the four NSF-funded supercomputer centers, whose funding is scheduled to end after fiscal 1997.

Last October, the National Science Board approved a continuation of funds through 1997, giving the task force two years to plan a strategy for fiscal 1998 and beyond. The task force committee will prepare an analysis with recommendations for the continuation, restructuring or phasing-out of the Cornell Theory Center, the National Center for Supercomputing Applications at University of Illinois at Urbana-Champaign, the Pittsburgh Supercomputing Center and the San Diego Supercomputer Center.

In the meantime, the centers received a modest increase in the president's 1996 budget proposal. The administration requested a 5.6% increase for the centers, from $70.9 million in 1995 to $74.9 million in 1996.

The task force is likely to consider some of the recommendations of the recently released National Research Council report on the High-Performance Computing and Communications program. Although the report broadly addressed the overall HPCC program, it also tackled issues involving the centers, suggesting the possibility of charging some users for access to computing cycles. The report also recommended focusing HPCC on computing and communications research areas and not classifying unrelated disciplinary activities at the centers under the HPCC umbrella.

Although most experts do not doubt the history of technological breakthroughs at the centers, most see a need to adapt to a changing technological and political environment. For many users, desktop workstations provide sufficient power. Industry has come to accept and experiment more with parallel machines. Political pressures have mounted to bring the centers in line with a broader social agenda, and federal budgets are becoming more constrained.

Other items on the committee's agenda include:

• How best to meet future needs of science and engineering communities for high-end computational resources.
• The appropriate role of NSF in driving leading technologies such as parallel and distributed computation into scientific and engineering applications; and NSF's role in interacting with computer scientists and engineers, computational mathematicians, vendors of parallel systems and industrial users.
• The appropriate range of potential grantees and suppliers in such a program, and the potential for leverage of NSF program funds by partnering with other federal and state agencies, technology vendors, universities and industrial users.
• The potential needs of high-end computational users versus those of more information-intensive users.
• Expected budget realities for the first five years of any recommended program.

The members of the committee are:

Ed Hayes (committee chair), Ohio State University

Arden Bement Jr., Purdue University

John Hennessey, Stanford University

John Ingram, Schlumber Limited

Peter Kollman, University of California at San Francisco

Mary Vernon, University of Wisconsin at Madison

Andy White, Los Alamos Advanced Computing Laboratory

William A. Wulf, University of Virginia.