Search
Sites and Affiliations
Leadership
Research and Applications
Major Accomplishments
Frequently Asked Questions
Knowledge and Technology Transfer
Calendar
Education and Outreach
Media Resources
Technical Reports and Publications
Parallel Computing Research - Our Quarterly Newsletter
Contact Information
CRPC Home Page

UniGuide Featured Site

The HPCC Advisory Committee: an Interview With Ken Kennedy 02.28.97

Source: HPCwire February 28, 1997
By Alan Beck
Editor in Chief


San Diego, Calif. -- HPCwire recently noted the appointment of Ken Kennedy, director of the NSF-funded Center for Research on Parallel Computation (CRPC), headquartered at Rice University, as co-chair of the new HPCC Advisory Committee (see article 10745, "PRES. CLINTON NAMES CO-CHAIR, MEMBERS OF HPCC COMMITTEE", 02.14.97). To learn more about the Committee's tasks and direction, HPCwire interviewed Kennedy. Following are selected excerpts from that discussion.
HPCwire: What priorities do you foresee the HPCC Advisory Committee will emphasize?

KENNEDY: "I believe the Federal Government will want us to spend a substantive amount of time thinking about the Next Generation Internet. It's high on their priority list, because they've budgeted around a hundred million dollars to start that effort. Presumably there will be additional funding for it in future years, assuming the original money is invested wisely. I'm sure Anita Jones, who chairs the Coordinating Committee for Communication, Computation, and Information Technology, also sees this as a high priority. And I also believe it requires immediate attention.

"One has to understand both the technical and investment strategy necessary to take such an investment and derive a real infrastructure from it. Much thought must be given to how to connect those places that can develop the applications enabled by the dramatically increased bandwidth of the Next Generation Internet. The current Internet succeeded because it connected the innovators, the government and research labs where experimentation was done and new applications developed.

"I believe similar applications can be supported on a much higher bandwidth Internet, but I believe there will also be important new applications developed. So we want to make sure that the policies undertaken as a result of our advice would encourage the development of next-generation "killer" apps for networking.

"There are technical issues as well regarding how the Federal Government should spend the limited amount of money that it has available for this project to connect all the relevant players: school districts and companies as well as research labs. There's a question of who should bear the cost and how it might be spread out over all the participants. However, we're not going to get too deeply involved in issues of regulatory policy."

HPCwire: Do you feel formation of the Advisory Committee heralds more governmental sensitivity to requirements for significant progress in HPC?

KENNEDY: "I have high hopes. I personally would like to make sure that, while appropriate attention is paid to networking and information technology, high-end computing is not forgotten in the rush. HPC is important for the country...Applications enabled at the high-end are vital for our quality of life. They make cars and airplanes safer, produce more effective medicines, and improve the general level of knowledge in our society. In my view, it would pay to get a five-year jump on addressing some of these critical applications.

"Funding for HPC has shifted and shrunk somewhat, although I believe there are still plenty of resources out there. For example, there are programs like ASCI -- because they need high performance computing; they need extensive numerical simulation. Programs like that -- and others will arise -- will look to high-end computing as very viable and important. But there is a squeeze on Federal funding for all programs, and computation is not exempted from this. Money is tight; many computer scientists have trouble getting grants, and I think that will continue. I do hope, though, that the Committee's recommendations -- considering its diverse membership -- will favor insuring that there is at least a maintenance and enhancement in these areas."

HPCwire: What role, if any, will the Committee play in the PACI competition and the reshaping of this country's HPC infrastructure?

KENNEDY: "We will not be involved in the selection of winners in programs like PACI or even in design of specific new programs. The Committee would be more likely to advise as to what kinds of competetively bid programs would make sense. This will be done on a fairly broad and abstract level. The Federal agencies' program managers are charged with the responsibilities of designing funding programs to carry out those agencies' goals. It is not our job to design Federal funding programs. But it is our job to advise on what kinds of programs make sense in order to move the technology forward, to make sure the research, corporate, and other communities we serve have the right technologies at the right time -- and hopefully ahead of the rest of the world, so we would continue to have the economic advantage. We will talk about what makes sense for the infrastructure of the country -- for example, the Next Generation Internet."

HPCwire: What kind of HPC technologies and infrastructure will make sense for us in the coming century?

KENNEDY: "One important issue that will be raised in this connection is whether we should continue to invest in development of computer architectures to advance to the petaflops level...What kinds of software challenges will be involved? Will the current workstation- or PC-based- scalable computing technologies be able to scale to a petaflops by 2007? There are some people who doubt it. This might well be an issue for a task force spun out of this committee."

HPCwire: How closely will the Committee examine security issues?

KENNEDY: "Security is an incredibly important issue and will grow in importance as the information infrastructure grows, especially in the commercial arena. Take the "Boeing product design problem", for example. Boeing designs airplanes not just in-house but with the help of hundreds of subcontractors, and they try to build these planes with paperless designs inside computers. But if you want a geographically dispersed team to explore such designs, you must have recourse to some kind of network. Currently, Boeing restricts their work to an internal network because of security concerns. However, if they found security strong enough, they might be willing to do this work on a global information infrastructure. The designers, located all over the world, might then use virtual-reality modeling to share and explore designs.

"Reliability will be as important as security. For example, if you're downloading extremely high-resolution video, and one link goes down, the performance of the entire network might be degraded. The whole area of high- content computing necessarily involves both reliability and security. That is another aspect of the current Federal program that we'll probably look at."

HPCwire: Will your Committee address issues involving export of encryption technologies?

KENNEDY: "We will stay away from regulatory issues involved with commerce. The policy of encryption export will be one that we will almost certainly duck. If you look at the composition of the Committee, I believe there's not a single lawyer on it. And legal issues are involved in that question. Now, if you polled the members, I would guess most are in favor of relaxing restrictions on export of encryption technologies. However, we will stick to longer-range views and try to help the Federal government assess what kinds of investments should be made over the next four or five years to insure continued technological leadership."


Alan Beck is editor in chief of HPCwire. Comments are always welcome and should be directed to editor@hpcwire.tgc.com
Copyright 1997 HPCwire. Redistribution of this article is forbidden by law without the expressed written consent of the publisher. For a free trial subscription to HPCwire, send e-mail to trial@hpcwire.tgc.com.