Need for supercomputer centers still exists

By Fred W. Weingarten
CRA Staff

NEWS ANALYSIS

The Task Force on the Future of the National Science Foundation Supercomputer Centers, chaired by Edward Hayes of Ohio State University, has been circulating a draft report guiding NSF on the renewal of grants for the centers.

The draft report was circulated for discussion, but reaction was fairly muted. Science magazine carried an article on it, and several researchers in the community commented on the document, which is still being refined. At press time, a final version still had not been presented for approval by the National Science Board (NSB). However, unless a crisis strikes, the report could well be approved by the time you read this article. The basic outline of the analysis and recommendations are clear, however, and unlikely to change significantly.

The purpose of the report is to guide NSF in dealing with the upcoming expiration of the current set of agreements with the four national centers. The centers have been operating for nearly 10 years and are costly investments. Even in the best of times, NSF--in particular, NSB--has been uncomfortable with large, open-ended commitments. And these are not the best of times. NSF's budget is likely to be flat or even shrink somewhat over the next five years. Congress has told the agency it should feel lucky to have done as well as that. In that environment, any programmatic response by NSF to National Supercomputer Center renewal will be carefully scrutinized. That usually means several studies and reports will be done.

In this case, the task force comes about two years after the Blue Ribbon Committee chaired by Lewis Branscomb of Harvard University. Although it may seem redundant to keep studying such questions as the renewal of the supercomputer centers, it is quite typical for major government policy decisions to be based on a sequence of such efforts. This is how decisions get made and how agencies can be protected from undue political second-guessing.

A lot of money and institutional prestige are at stake. Although NSF decisions have been relatively immune from political considerations, there will no doubt be pressure brought to bear on any actions regarding either a recompetition or a renewal of funding. Many decision makers in government have to be convinced to go down a particular path, and because public money is being spent, a solid rationale and a careful record of deliberation has to be established. It is hoped that in the process, some consensus of the community can be found. I believe this report captures this with three common-sense observations:

• The supercomputer centers are still valuable resources.
• Given 10 years of change in the computational research environment since the first centers were established, the program needs to be refocused to reflect current realities.
• There needs to be a recompetition because of the refocus and because no research program should be automatically renewed without having to undergo an occasional peer-review competition.

As requested in the NSF director's charter for the task force, the report focuses much of its attention on defining the missions of the supercomputer centers. Their missions, at least in congressional eyes, always have been a blur of possibilities, drifting with the political climate. Missions have ranged from providing test beds for the supercomputer industry to developing software for Internet applications, from research in computer graphics to serving as cycle centers for scientific users.

The task force redefined the centers around research in computational science and research in support of the evolution of the National Information Infrastructure (NII), with some service for very-high-end users still provided. The report also emphasized the need to move away from focusing on providing classical vector-processor cycles and toward access to alternative architectures, especially scalable parallel systems.

In recommending recompetition of the centers, the task force stood in contrast with the Branscomb report, which recommended renewal without recompetition. NSB rejected that recommendation, which likely was never politically salable for reasons stated above.

The task force did recommend that the recompetition be conducted under a revised format that stresses linkages among the centers and other smaller specialized centers of research in computational science and NII-related issues. The grants should be for five years, renewable for another five years.

Computer science and engineering is hardly mentioned in the task force report. This is not unexpected because it is, after all, a report about computer centers and computational science. So what does this have to do with computer science and engineering? Plenty.

There are broad implications and challenges for the CS&E community, especially if the recommendations as presented in the draft are accepted by NSB and implemented by NSF. There is an interdependence of interests between the computational science and the CS&E communities. Further, computer science and engineering will play a major role in creating the NII, not just in advancing the fundamental underpinnings of high-speed data communication, but in pushing the fundamental science underlying major social applications such as educational technology and digital libraries.

I believe that, should NSB approve the task force recommendations, an immediate competition for renewal of the centers will begin. Inevitably, it will be a lengthy evaluation process, and time is running out before the existing grants expire. The competition would most likely be around the framework provided by the report: national supercomputer centers with a stronger research agenda and links with smaller, specialized research centers, some of which could have a strong computer science or computer engineering focus.

If this occurs, this will present an opportunity for the CS&E and computational science communities to begin searching for some common ground and new collaborations that will benefit both communities.