NRC: Is scope of HPCC too broad?

By Juan Antonio Osuna
CRA Staff

In recent years, computing researchers have surfed the wave of public enthusiasm for information technology, allowing their activities to be subsumed under the politically popular High-Performance Computing and Communications Initiative (HPCCI). But what happens when this wave hits the shore and the program loses its political momentum? Will Congress and the executive branch continue to see computing research as a vital social concern?

The last question was suggested by a report released in February by the National Research Council, titled Evolving the High-Performance Computing and Communications Initiative to Support the Nation's Information Infrastructure.

"Today the HPCCI supports nearly all [computing] research, an arrangement that is both misleading and dangerous: misleading because much important computing research addresses areas other than high performance, and dangerous because reduced funding for the HPCCI could cripple all of computing research," the report said.

Produced by NRC's Computer Science and Telecommunications Board, the report issued broad and detailed recommendations for guiding the future of HPCC. On a broad level, it warned against putting all computing research efforts under one label and recommended greater political diversification.

"The 'war on cancer' did not support all of biomedical research, and neither should the HPCCI or any future initiative on the nation's information infrastructure subsume all of computing research," the report said.

This recommendation takes on special significance in the current climate of political uncertainty. Last year, Republicans placed the HPCC program on their list of possible budget cuts, which indicated a $1.2 billion reduction in HPCC over five years. Yet House Speaker Newt Gingrich has voiced overall enthusiasm for a government role in information technology. It is too early to predict how HPCC will fare under the new Congress.

The report's most prominent message to Congress may be that computing research has become essential to social and economic development and will only become more so.

"Computing research continues to be dramatically undersupported compared to its importance to society and its contributions to the economy," said CRA Board member Ed Lazowska, who served on the committee that prepared the report.

The report enumerated many technologies that were developed under government funds and then moved quickly into the commercial world. The most publicized example may be the National Center for Supercomputing Applications (NCSA) Mosaic Web browser.

Since Mosaic's inception, the number of Web servers has increased 100-fold, and more than 1 million copies of the Mosaic software have been downloaded. The NCSA-developed software has spawned massive development in the commercial world, which now offers 20 licensed versions of Mosaic and has created more than 20 million licensed copies. The software has become so popular that Microsoft Corp. plans to package Mosaic with its Windows 95 operating system.

While much of the report touted the benefits of computing research and made a case for increased--or at least sustained--funding, it also offered a critical analysis of how the HPCC program can be improved and shaped.

For instance, the report echoed concerns expressed by a General Accounting Office report released in November that said software has lagged behind hardware development on parallel computers. Specifically, the NRC report said research should focus on developing better compilers and programming languages with improved portability across machines of different sizes.

NRC's report also warned against wasteful spending in areas best left to industry. "Avoid funding the transfer ['porting'] of existing commercial applications to new parallel computing machines unless there is a specific research need," the report said, explaining that such transfer does little to expand scientific knowledge because most of these applications were written for sequential or vector machines.

Government funding has done much to establish parallel hardware as a viable tool in the commercial world. Now that it is established, the report said, the HPCC program should refrain from funding "industrial stimulus" purchases of hardware and the development of commercial hardware by computer vendors.

In the area of networking, the report urged a greater focus on reliability and performance of large-scale networks offering distributed information systems and on network applications that, like Mosaic, increase network access.

Many of the report's recommendations concern the four National Science Foundation-managed supercomputing centers, which constitute the largest chunk of the entire HPCC program. Much of supercomputing money is inappropriately classified under HPCC, the report said. "Use of HPCCI funds is appropriate only when the research contributes significantly to the development of new high-performance computing and communications hardware or software," the report said.

The centers serve the routine needs of scientists in various disciplines by granting free access to computing cycles. To continue these services outside HPCC, NSF should consider "charging mechanisms" to users and additional non-HPCC funding sources within disciplinary directorates.

As for overall management of the program, the report recommended that Congress establish an advisory committee intended to provide broad-based, active input to HPCC. The government also should appoint a full-time coordinator to serve as the program spokesperson and advocate for HPCC.

To ensure that long-term research goals are not too dependent on the HPCC program, the report concluded that federal funding agencies should identify research areas that are long-term and independent of HPCC.

"This problem is particularly acute at NSF, where nearly all of the funding in the Computer and Information Science and Engineering Directorate is coded as HPCCI funding," the report said. "Given that NSF is not a mission agency, this approach seems shortsighted. Ongoing funding of important research areas in computer science will be critical to the nation's future, independent of the future of HPCCI."