Source: HPCWire, August 21, 1998
Commentary by Norris Parker Smith, editor at large

The following phrases and exhortations should stir familiar memories among members of the high-performance computing community who have been following US federal policy on supercomputing since the 1980s:

--"The current scalable parallel high-end computing systems are not well suited to many applications of strategic importance to the nation. ...funding should be focused on innovative architectures, hardware technologies, and software strategies that overcome the limitation of today's systems."

--"...federal support for research in information technology is dangerously inadequate."

--"The current federal program is inadequate to start necessary new centers and research programs."

--..."increases in research on software should be given the highest priority."

--"[The goal should be] sustained petaop performance on real applications by 2010."

NEW TERMINOLOGY, OLD GOALS

There is, to be sure, some unfamiliar terminology:

--"High-end computing" has displaced supercomputing and high-performance computing.

--"IT" has been consecrated at last by the American bureaucracy. This is a timely change, because IT (for Information Technology] has long been absorbed into general usage elsewhere in the world.

--And, of course, the goal of high-end computing has been escalated from teraflops to petaops.

These observations and recommendations issue from an advisory committee that is both new and old.

The name is new: President's Information Technology Advisory Committee (PITAC), conferred by the President on July 24. He retitled the existing Advisory Committee on High Performance Computing and Communications, Information Technology, and the Internet -- clearly a positive step forward.

It submitted an interim report to the President on August 10. The executive summary was transmitted by HPCwire (see HPCwire [ ] 13656); greater detail can be found on the Web at http://www.ccic.gov

PITAC, like its antecedents, is composed mostly of persons with a long-standing professional and personal connection to IT -- and, in many cases, to high-end computing.

Its co-chairs are Bill Joy of Sun Microsystems, and Ken Kennedy of Rice University.

Members include Eric Benhamou (CEO of 3Com), Vinton Cerf (now with MCI), Danny Hillis (Walt Disney), Larry Smarr (NCSA) David Nagel (president, AT&T Labs), Leslie Valdez (Intel), James Gray (Microsoft) Steve Wallach (CenterPoint ventures), and Irving Wladawsky-Berger (IBM).

The roster also includes representatives of Simmons College, the University of Pennsylvania, the University of Washington, Stanford, Berkeley, Montana State University, Mississippi State University,and Carnegie-Mellon.

A USEFUL FUNCTION, BUT ...

It should be recognized that committees of this kind serve useful purposes. They help to define agendas for the White House and Congress. They seek to represent the interests and aspirations of the communities from which they are drawn.

And it should be remembered that similar committees serve similar purposes for the medical community (and, indeed, the numerous fragments thereof,) agriculture, road-builders, conservationists, and a long list of other interests that have an indisputable right to make their voices heard in Washington.

Joy and Kennedy are highly-regarded, thoughtful, and experienced, excellent choices for a body like PITAC. The committee members that this reporter knows are all highly-qualified, and the other members are presumably of similar accomplishments.

It should also be recognized that the problems -- and the financial appetites -- of the IT community have not changed fundamentally during the past two decades. Despite all the strides forward and upward, performance is never sufficient, and there is never enough money.

Nevertheless, it seems reasonable to expect that the enormous changes that have taken place during these decades would form the basis for a less conventional agenda.

The executive summary of the PITAC report notes that IT businesses "have accounted for one-third of the total growth in U.S. production since 1992, creating millions of high-paying new jobs."

This represents an enormous investment by the firms responsible for this growth. Companies like 3Com, MCI, Hughes, SGI, Microsoft, Intel, Lucent, Walt Disney, and IBM -- all represented on the PITAC membership -- expend a total of many billions of dollars on R&D each year.

Presumably, a large proportion of their R&D programs are devoted to higher performance at the high end, better software, and other objectives advocated by PITAC. Federal expenditures would be marginal and could well be misplaced.

PITAC's report also notes that "Computers on university campuses and other civilian research facilities are falling rapidly behind the state of the art." It also calls for more "federal research on the social and economic impacts" of IT and "government/university/industry partnerships to increase IT literacy, access, and research capabilities."

These are praiseworthy goals, but they seem like footnotes to the main thrust of the PITAC recommendations.

REDUCING DIFFUSION

Finally, PITAC notes that federal IT research is diffused among a number of agencies. As funding has dwindled, each has focused more sharply on their own mission-oriented goals, leading to excessive attention to short-term research and a neglect of fundamental research.

PITAC advocates a lead federal agency for coordinating IT research and nominates the National Science Foundation as "the most logical choice."

This is a sensible suggestion, but almost certainly doomed. Traditionally, successful efforts to increase or sustain funding for HPC have depended upon the establishment of carefully-crafted coalitions of government agencies (and their respective constituencies).

These compromises are less than ideal, but they reflect the realities of doing business (and securing financial support) in Washington.

Many years ago, this reporter spent some time in the foreign aid business. The hardest part was persuading recipient countries that they had progressed so far that they didn't need foreign money any more.

Indeed, after the checks stopped arriving from Washington, the graduated recipients discovered that they did better on their own. If carried on too long, the habit of dependency can be very corrosive.

Has high-end computing in the United States -- and IT as a whole -- qualified for graduation, or for stipends concentrated on new objectives?

This issue certainly deserves examination by PITAC.

Norris Parker Smith, scrib@well.com, is a journalist who specializes in HPC and high bandwidth communications. Reader comments are welcome.