3. Program Overview
The best source for getting a current overview of the HPCC Program is
the 1994 document, "High Performance Computing & Communications:
Toward a National Information Infrastructure" [6]. For purposes of
explaining details of the HPCC Implementation Plan, portions of that
material are included here. This section provides a broad overview
of the definition, organization, and management used to implement the
HPCC Program. The HPCC goals and objectives provide the focus for
the global management, planning and organization of activities.
Through a process of collaborative coordination, the participating
agencies seek to leverage each other's activities wherever possible
and to minimize redundancy in programs. However, funding for the
HPCC Program flows directly to each agency participating in the
Program. Each agency has its own mechanisms and approach to select
and evaluate projects funded under this Program. A wide variety of
methods are used to distribute the results of HPCC activities and to
evaluate the overall progress on the HPCC Program.
3.1. Goals and Objectives
The High Performance Computing and Communications Program goals are to:
o Extend U. S. technological leadership in high performance
computing and computer communications;
o Provide wide dissemination and application of the technologies
to speed the pace of innovation and to improve the national
economic competitiveness, national security, education, health
care, and the global environment; and
o Provide key enabling technologies for the National Information
Infrastructure (NII) and demonstrate selected NII applications.
These goals will be realized by accelerating and focusing the HPCC
activities within each of the federal agencies charged with carrying
out this program and coordinating those activities among the
participating agencies. Many critical applications in industry,
national security, and the environment require far greater computing
capability than is currently available. The HPCC Program will
accelerate the development of scalable computing systems that will
have the capability and capacity to address more of these critical
applications. It will also accelerate development of the critical supporting
technologies, such as file storage systems, computing environments
and network communications required to support the effective use of
these computers. The Administration's vision for a National
Information Infrastructure places unprecedented demands for network
connectivity, capacity, database availability, information
management, access security, and ease of use. The HPCC Program will
work with industry to help to create important elements of the
technology base needed for a universally accessible NII and will use
this technology to develop and demonstrate prototype "National
Challenge" applications. All of these activities depend heavily on the
development of more cost-effective approaches to writing and
maintaining software and the development of a larger pool of
highly-trained professionals to use these computers and networks.
The HPCC Program will develop and deploy software, tools, and
improved algorithms to meet these programmatic needs. It will also
establish sufficient human resources of educators, scientists, and
other trained professionals in computational science and engineering
to permit effective use and application of the HPCC technologies.
The HPCC Program is organized into five components with the following
key aspects:
High Performance Computing Systems (HPCS):
o Accelerated development of scalable computing systems, with
associated software, including networks of heterogeneous systems
ranging from affordable workstations to large scale high
performance systems
o Technology advances in component packaging, mass storage, and
communications that support the design and use of scalable
computing systems
National Research and Education Network (NREN):
o Broadened network connectivity of the research and education
communities to high performance computing and research resources
o Accelerated research, development, and deployment of networking
technologies
Advanced Software Technology and Algorithms (ASTA):
o Prototype solutions to Grand Challenge problems
o Improved algorithms, software technologies, and software tools
for more efficient use of scalable computing systems
o Deployment of advanced high performance computing systems for R&D
Information Infrastructure Technology and Applications (IITA):
o Prototype solutions to National Challenge problems using HPCC
enabling technologies
o Accelerated development and deployment of NII enabling technologies
Basic Research and Human Resources (BRHR):
o Support for research, training, and education in computer science,
computer engineering, and computational science; and infrastructure
enhancement through the addition of HPCC resources
3.2. HPCC Management, Planning and Organization
The High Performance Computing and Communications Program is
implemented as a partnership among federal agencies and other
organizations. The HPCC Program reports to the Director of the
Office of Science and Technology Policy (OSTP). At the direction of
the Director, Program oversight and budgetary review are provided by
the Committee on Information and Communication (CIC) of the National
Science and Technology Council. The National Coordination Office
(NCO) for High Performance Computing and Communications provides a
central focus for the Program. The Office coordinates the activities
of the agencies, other organizations participating in the Program,
and organizations providing services to it. The Office also provides
an interface to Congress, industry, academia, and the public. Donald
A. B. Lindberg, the NCO Director (and concurrently Director of the
National Library of Medicine), reports to John H. Gibbons, the
Director of the Office of Science and Technology Policy.
The High Performance Computing, Communications, and Information
Technology (HPCCIT) Subcommittee handles planning for the HPCC
Program. The Subcommittee meets monthly to coordinate agency HPCC
Programs through information exchanges, the common development of
interagency programs, and the review of individual agency plans and
budgets. HPCCIT charters a variety of Working Groups to coordinate
activities in specific areas, Task Groups to carry out specific
limited duration studies, and Task Forces to make various planning
recommendations. In most cases, individual agencies are assigned
responsibility to lead these groups. Currently active groups
include:
The Scientific and Engineering Computing Working Group, led
by NASA, has the mission to identify, promote, transfer,
and coordinate software developed under HPCC.
The NREN Group, led by NSF, coordinates network integration
activities and works closely with the Federal Networking
Council (FNC), which consists of representatives from
federal agencies that use federally funded networks. The
FNC is responsible for coordinating the efforts of
government HPCC participants and other NREN governmental
constituents, in addition to providing a liaison to
nonfederal communities interested in the federal program.
A Federal Networking Council Advisory Committee of
nonfederal scientists and network users has been created to
support the FNC.
The IITA Task Group, originally led by NSF, was chartered
to acquire information on IITA-related activities planned
and in progress within the HPCC agencies, to develop a plan
for coordinating IITA activities among those agencies, to
assist HPCCIT in reporting requirements related to IITA,
and to provide advice on technology and policy related to
IITA to HPCCIT and its parent committees. This group
continues under the current leadership of ARPA.
The Task Force on Working Groups is evaluating the current
structure and coverage of HPCCIT Working Groups and will
make suggestions for changes in modes of operation and
coverage.
3.3. Selecting Projects for Funding
To reduce redundancy and encourage adequate funding of all HPCC
areas, the HPCCIT Subcommittee has worked with the agencies to define
complementary areas of responsibility. Within each agency, the
actual mechanisms used to evaluate and fund activities differ.
Several agencies conduct R&D by agency staff (and/or national
laboratories), while others work through communities of researchers
in academia and industry. Each agency that funds external R&D
activities has a program to review and fund competitive, merit-based
awards that are consistent with agency mission. Calls for proposals
for these grants receive wide distribution, including both electronic
means (e.g., electronic bulletin boards and Mosaic servers on
Internet) and traditional hard copy (e.g., Commerce Business Daily).
3.4. Progress Measures
The primary method of measuring progress of the Program is the
establishment and subsequent review of yearly milestones. The FY
1994 milestones for the overall Program can be found in Section 4 of
this document. These milestones are firm expectations, based on
funding that has been estimated for this fiscal year. The FY 1995
milestones (also shown in Section 4) describe tentative expectations
for the Program, based on the anticipated funding at the President's
requested level. Achievement of these high level milestones depends
on specific program activities that take place within each agency.
Section 5 gives a more detailed look at each program activity,
including their expected milestones. The combined successful
completion of these activity-specific milestones across all agencies
translates into meeting the various HPCC Program milestones.
The nature of these milestones (and hence the definition of
"measuring progress") depends on the activity undertaken. In some
cases, clear quantifiable measures accurately convey progress (e.g.,
the number of colleges and universities connected to NREN, the speed
of network communications, and the computational time needed to solve
a specific application problem). In these cases, the HPCC Program
and Project Activity milestones use these quantifiable measures as
the primary measure of progress. In other cases, a number of
different quantifiable measures have been proposed, but their use
does not always accurately reflect progress. For example, developing
a new computer system with a peak execution speed of 100 gigaflops
may not be progress if that rate can only be achieved on artificially
constructed test cases. In these cases of uncertain quantifiable measures,
a critical aspect of the HPCC research activity is to identify and
define useful measures. Finally, in cases when quantifiable measures
are unavailable or not necessarily reflective of the progress
achieved, the HPCC Program relies on qualitative measures. The HPCC
Program combines the advice and recommendations of technical
reviewers and advisory committees with program managers' reviews to
assess progress.
3.5. Reporting Results and Interactions
The HPCC Program uses a variety of vehicles to convey its results.
At the individual projects level, investigators present their results
in workshops, conferences, and journal publications. In addition,
investigators rely on electronic means (such as the HPCC Software
Exchange and file transfer services) to quickly share and distribute
their tools. At the agency level, periodic reports summarize the
results of the HPCC Program [7]. In addition, the HPCC agencies are
beginning to establish electronic information servers to provide
direct access to the results of their research projects. At the HPCC
Program level, the National Coordination Office publishes a yearly
report that summarizes the highlights of the overall Program. The
NCO has also established its own information server that is being
linked to all of the agency-level servers. When completed, anyone
with access to the Internet will be able to access technical results
throughout the HPCC Program.
Outreach conferences and workshops play an important role in
reporting results of the HPCC Program and receiving feedback on
progress. High Performance Computing and Communications: Technology
for the National Information Infrastructure [8] identifies six
HPCC-sponsored conferences and workshops as well as 13 other ones in
which HPCC played a prominent role. Additional events (sponsored
and/or involving HPCC) since that report include:
o Workshop on Enabling Technologies for Peta(FL)OPS Computing
Systems (February 1994),
o R&D for the NII: Technical Challenges (February 1994), and
o ARPA HPCC Symposium (March 1994).
In addition to conferences and workshops, representatives from the
NCO and HPCC agencies meet with and provide briefings to companies,
universities, associations, media, and other groups. In 1993, the
NCO met with over 80 such groups.
3.6. Program Oversight
The High Performance Computing Act of 1991 (Public Law. 102-194)
requires the President to select a High Performance Computing
Advisory Committee of nonfederal members, including representatives
of the research, education, and library communities, network
providers, and industry, who are specially qualified to provide
advice and information on high performance computing. While this
committee has not yet been appointed, representatives of industrial
companies, organizations, and associations have met with the HPCCIT
in public sessions to advise the agencies. Detailed program reviews
are also carried out by individual agency via their management review
structures, including official advisory committees. The NREN especially
has benefited by advice of the Federal Networking Council, which is
appointed by the Director of NSF. In addition, both CIC and Congress
have oversight responsibility for the Program. This section
highlights oversight and review activities that have been carried out
recently and describes existing review mechanisms within each of the
agencies.
The HPCCIT Subcommittee encourages broad comment and external
recommendations on the HPCC Program. For example, the Subcommittee
has devoted portions of its meetings to hosting visitors. Three
recent meetings included representatives from computer system
vendors, the telecommunications industry, and software industry.
o The November 1993 meeting included 17 representatives
from computer system vendors and consortia. One important
message that came out of this meeting was the recommendation
to push for more involvement of independent software vendors
to get critical applications moved to the emerging massively
parallel platforms.
o The January 1994 meeting included 15 representatives from
the telecommunications industry and related forums. Several
key messages recurred throughout this meeting: (1) a desire
by a broad range of companies to participate in providing
the information superhighway to the public, (2) the key
role of the federal government in working with industry to
establish telecommunication standards, and (3) the need for
a level playing field (in reference to government regulations)
to provide new capabilities to customers.
o The February 1994 meeting included 18 software industry
representatives. This meeting produced a much wider
spectrum of opinions and issues than the previous two.
Commonly mentioned points included: (1) the common base
of software development tools must be improved, (2) third
party software vendors are reluctant to enter the HPCC market
due to its small size and the relatively high development costs,
(3) better protection is needed for software as intellectual
property and corporate assets, and (4) National Challenge
applications that are close to market needs and generate
substantial market interest should be stressed.
Similar HPCCIT Subcommittee meetings are in the planning stages to
acquire feedback from academia. To further the feedback process from
these meetings, the HPCCIT Subcommittee will actively seek
participants' comments on this Implementation Plan.
A variety of reviews, studies, and hearings on various aspects of the
HPCC Program have taken place in the past 15 months and others are in
the planning stages. They include:
o Computer Science and Technology Board of the National Academy
of Science commissioned study of the federal HPCC Program
(due Summer 1995)
o General Accounting Office study of the ARPA, NSF, DOE, and
NASA HPCC Programs. (in progress)
o The NSF report to the Senate Appropriations Committee entitled
"Overview of the Management of the NSF HPCC Program" (March 1994)
o Informal briefing for Committee on Information and Communication
(March 1994)
o Congressional Staff briefing (February 1994)
o Congressional Staff briefing (September 1993)
o The National Science Board commissioned Blue Ribbon Panel Report
on the NSF HPC program [9]
o Testimony before Senate Committee on Commerce, Science and
Transportation, Science Subcommittee (August 1993)
o Congressional Staff briefing (July 1993)
o Congressional Budget Office study of the HPCC Program [10]
o Testimony before House Committee on Science, Space and Technology,
Subcommittee on Science (May 1993)
o A General Accounting Office Review of ARPA's HPCC Program [11]
o Testimony before House Committee on Science, Space and Technology,
Subcommittee on Technology, Environment, and Aviation (March 1993)
o Testimony before House Committee on Energy and Finance, Subcommittee
on Telecommunications and Finance (February 1993)
o Congressional staff briefing (February 1993)
The ARPA reviews its HPCC efforts at many different levels to insure
consistent program evaluation in a dynamic R&D environment. Projects
produce reports on a regular basis that are reviewed by program
managers. An annual process of updating accomplishments milestones
and project plans is tied to the incremental funding process. Site
visits, project meetings, principle investigator meetings, and
regular interactions using the Internet as an information
infrastructure enable ARPA staff to keep up with their program
management responsibilities. In addition, contracting details are
handled by contracting agents that work with ARPA as part of the
program management process. Office directors and program managers
develop plans and milestones that are approved by senior management
during the planning and budget cycle. New programs and ideas are
proposed during this process. In conjunction with yearly funding
decisions by ARPA and DoD, senior ARPA technical management
critically review program areas, plans and accomplishments. Guidance
is provided to reflect programmatic, technical and funding
directions. At the DoD level, programs are described through a
formal process that requires Agency, ARPA Comptroller, DoD
Comptroller, senior ARPA management, and senior DoD approvals. Once
approved, these descriptions become part of the Defense budget
submitted to Congress for approval. In addition to other internal
Federal HPCC reviews, there are Congressional Briefings, High
Performance Computing Agency crosscuts, technical working groups, DoD
advisory panels, and a number of National Academy of Science studies
which strongly contribute to the planning process.
Numerous processes and mechanisms are in place that contribute to the
definition and review of the NSF HPCC program long term goals and
objectives. In addition, more specific goals and objectives are
established for each of the activities within the program. The
National Science Board, panels and committees commissioned to study
and recommend program activities, and external advisory committees
contribute primarily to the more general, long-term goals. The
establishment of goals, objectives, implementation mechanisms,
evaluative measures and the broader value of specific activities is
accomplished through such means as: external peer review (mail
reviews, panel reviews, and site visits), workshops for developing
research agendas, program committees of visitors, technical oversight
teams, ongoing site visits by program staff and outside experts,
program officer review of final project reports, and the bodies of
opinion held by the community of researchers themselves. Evaluation
of progress is integrated into the management process as appropriate
to each class of HPCC activities. Examples are:
o Blue Ribbon Panel for the NSF HPC program
o Federal Networking Council Advisory Committee for parts
of the NSFNET
o Directorate Advisory Committees
o Committee of Visitors for each research program
The DOE HPCC program focuses on basic mathematics and computational
research and on developing and delivering technology for use by other
scientists and engineers in DOE and US industry. Performance
evaluation is an integral part of these programs. Because of this
focus, external review by prospective users of the technology is a
critical component of measuring performance. In previous years this
has been most explicitly present in the use of committees of users
such as the ESnet Steering Committee and the ER Supercomputer Users
Group to evaluate the effectiveness of the access and networking
programs. Many of the education programs established under the DOE
HPCC program have evaluation built into them. In the area of
software technology, using prospective users of technologies as
reviewers has ensured that the technology developed is desired by
users. In FY 1995 and beyond, DOE will further formalize these
procedures to include program-wide reviews of the basic technology
components of the program by significant prospective users of those
technologies. These will include reviews of effectiveness in all
categories as well as specific numerical targets.
The NASA HPCC program evaluations take place at several levels. At
the agency level, the NASA Advisory Council (NAC) has established the
Ad Hoc Task Force on Supercomputing, which has recently completed a
review and report on the NASA HPCC program. It is expected that
other such bodies will be permanently chartered under the Aeronautics
Advisory Committee to advise the NASA HPCC program. Within the
program, comprehensive reviews are conducted for each of the projects
annually. In addition to appropriate NASA personnel, representatives
from other federal agencies, academia, and industry may be invited to
participate. Annual reviews of program progress and plans will also
be conducted by the NASA HPCC Working Group. In addition, quarterly
reviews of the NASA HPCC program are conducted by program managers,
the associate administrator of aeronautics and the director of the
NASA HPCC Office.
The NIH HPCC program goals are enhancements of existing NIH program
missions to support biomedical science and expand biomedical
knowledge. Program objectives are developed by Institute Directors,
advisory bodies and senior program staff, and are peer reviewed for
determination of merit. Each of the participating NIH components has
one or more standing external advisory committees which review new
and existing programs. These include the National Library of
Medicine (NLM) Board of Regents, NLM Board of Scientific Counselors,
National center for Research Resources (NCRR) Research Resources
Advisory Council, NCRR Biomedical Research Technology Review
Committee, the National Cancer Advisory Board, the National Cancer
Institute Division of Cancer Biology and Diagnosis Board of
Scientific Counselors, and the Division of Computer Research and
Technology Advisory Council. The final decision regarding individual
HPCC programs within each of the participating ICDs rests with the
Director of that Institute. Within each of the participating NIH
Institutes, mechanisms exist to ensure objective evaluation of progress
and results and identification of remaining obstacles.
The National Security Agency reviews its HPCC support efforts on a
yearly basis in several separate reviews. A steering group comprised
of senior managers from the technical components provides high level
guidance prior to the formal budget process. The steering group
receives individual project assessments from the project managers and
determines whether any major shifts or changes are needed to the NSA
program. Senior management has the flexibility to sponsor HPCC
efforts in several budget reviews. Individual projects are proposed
and budgeted within the technical components. They are constantly
being evaluated by the project managers. Monthly status reports are
evaluated and meetings held with the project staff to ensure that the
correct focus is maintained. During the budget review cycle, the
projects are evaluated and retained based on their performance,
importance relative to other initiatives, and priority based on the
steering group guidance. New projects can be proposed by the
technical components each year during the NSA Technology Program
review process.
Annually NIST has the National Academy of Science review all the
programs and activities of each operating unit, as part of its normal
operation. Assessment includes relevance, performance measures and
achievements. Consequently, programs and activities that are a part
of the federal HPCC Program are subject to review and comments of an
external panel of experts from academia and industry. A written
report is presented to the Director of NIST, the Administration and
Congress. Selection of individual projects and subsequent progress
reviews are conducted by the program manager. These are reviewed by
an inter-operating unit panel comprised of senior managers for
relevance to agency mission, the program, and for continued
acceptable performance.
The NOAA HPCC program supports and enhances NOAA programs in
environmental prediction and stewardship. Comprehensive NOAA science
reviews of these programs are held periodically, with a review cycle
completed as of April 1994. Quarterly reviews of NOAA HPCC program
progress are conducted as an integral part of NOAA-wide quarterly
reviews by the NOAA Administrator and other NOAA senior line and
program managers. Overall NOAA HPCC program goals and plans are
reviewed annually as part of the NOAA strategic planning process.
The EPA HPCC program is focused on incorporating advances in
computing and communications technology into critical environmental
assessment applications and transferring those advanced tools to key
State, Federal, and industrial users. EPA senior management review
the EPA HPCC program annually to assess its relevance to the agency
mission and program achievements. Agency guidelines require an
external peer review of the EPA HPCC program every two years. The
external review panel is composed of representatives from other
federal agencies, academia, and industry. The first EPA HPCC
external review is scheduled for summer 1994. Within the program,
each major project is reviewed at least twice a year to evaluate
progress toward the program objectives.
3.7 HPCC Planning Beyond FY 96
Computing and communications are ongoing, strategic enabling
technologies. The current HPCC Program provides additional
stimulation to accelerate the progress of these technologies and the
benefits that accrue from their use. This Program was initiated as a
five year effort, covering FY 1992 through FY 1996. Since in-depth
planning for the FY 1997 budget will begin in the Spring of 1995,
preparations for that planning must begin now. Two avenues of
activity need to be pursued: a complete external review of the HPCC
Program to ascertain its likelihood of meeting and/or exceeding its
goals and an assessment of options and benefits that could accrue
from continued efforts to stimulate the technology. Given the recent
request to the Computer Science and Technology Board to perform an
HPCC Program review due next year, this activity would presumably
address a key part of the first avenue. In addition, the HPCCIT
Subcommittee will continue to seek input and comment from industry,
academia, and other segments of the government at its meetings. To
further the feedback process from these meetings, the HPCCIT
Subcommittee will actively seek participants' comments on this
Implementation Plan.
The assessment of options and benefits that could accrue from
continued efforts to stimulate computing and communications
technology has already begun. HPCC will continue to sponsor a
variety of workshops with technical experts, engage discussions with
potential users of the technologies, and prioritize critical directions
for future activities. Three HPCC workshops are a part of that process:
o Enabling Technologies for Peta(FL)OPS Computing Systems
(February 1994) sponsored by NASA, NSF, NSA, Ballistic
Missile Defense Organization, ARPA, and DOE;
o ARPA HPCC Symposium (March 1994); and
o System Software Tools for High Performance Computing
Environments (scheduled for September 1994).
The key questions that will be addressed throughout the planning process
include:
o What technical objectives in computing and communications
show the greatest promise for benefiting the public and
enabling U.S. industrial competitiveness with additional
stimulation?
o If successful, what results would accrue and who would
benefit?
o What are the expected costs and risks associated with the
objectives?
o Which agencies are most qualified to contribute toward
these objectives?
o What is the right mix of industry/academia/government
participation?
Through broad input and extensive discussions, the HPCCIT Subcommittee
expects to develop a plan and detailed proposals that will merge naturally
into the budgetary planning process.