Final report
The final report on high-end
computing in Japan is available
in PDF format. ( ~ 11.7 MB )
Individual chapters are also available.
Workshop
Viewgraphs presented at the May
25, 2004 workshop are available below
Scope
While the Earth Simulator
itself is fairly well known, the main task
for this study is to determine what comes after the Earth Simulator in
Japan.
The WTEC panelists will address this task by gathering information and
analyses
from a variety of sources: (a) their own existing knowledge of and
insights
into U.S. and Japanese HEC R&D activities; (b) a review of the
published
literature and any "gray literature" reports that can be found on the
subject; (c)
correspondence with U.S. and Japanese colleagues, as part of a careful
process of preparation for the Japan site visits; (d) a week of site
visits in
Japan, following several months of careful preparations for those
visits; (e) a
workshop at which the panelists present their preliminary findings and
receive comments on those findings from sponsors, U.S. colleagues, and
invited
Japanese site visit hosts and, (a) reviews of the panel's draft site
reports and
draft final report by the sponsors and Japanese hosts. After collecting
all
this information, the WTEC expert panelists add their own analyses and
report their findings back to the U.S. government sponsors in the most
timely manner
possible. The details of the study scope will be agreed on at the study
kickoff meeting, in consultation with both panelists and sponsors. A
preliminary list of issues to be covered might include the following:
Technical Issues:
- Review of likely future
Japanese machines, and of long-term
Japanese
research activities in the field such as the machine Fujitsu is
building for the
Japanese National Aerospace Laboratory (NAL).
- What innovations in
architecture, hardware, and software systems
are
under development for the next generation of Japanese high end
computing
systems? NEC's R&D and special architectural approaches in
supercomputers played a key role to the success of the Earth Simulator
(ES). What are the real,
perhaps under-publicized, technical innovations at NEC and other
Japanese
supercomputer vendors? What are their current research themes? What are
the technical barriers that NEC and other Japanese vendors feel must be
overcome to
achieve the next level of performance?
- Overall comparative
assessment of Japanese high end computing
research, development, and deployment programs relative to those in the
U.S.
- Efforts in the software
area, including open source operating
systems
and programming environments (e.g., Linux) and models of software
support
for
architectures aimed at achieving high memory-to-CPU bandwidth. What is
the Japanese model for developing and maintaining software
capabilities?
Topics to be covered in this area could include programming languages
and
environments, compilers, debuggers, libraries, and run-time tools.
- Conception, design, and
implementation of the Earth Simulator.
The
research and development plan for ES was openly publicized and well
known to the
international community. Our fact-finding mission will focus on how the
ES team executed the plan, what technical hurdles were encountered, and
how
they were overcome. What was the strategy behind decisions that were
made
concerning Internet access, data storage availability, etc.?
- Implications of the Earth
Simulator and likely follow-on machines
for
other competing architectural designs and machines in Japan (and the
U.S.),
including, for example, cluster systems. With the NEC success, are
companies like
Hitachi and Fujitsu likely following suit, or are they moving in other
directions?
- How do the Japanese evaluate
total cost of ownership of these
high end computing systems, in terms of how efficiently they are using
scientists' time,
dollars per code execution, etc.? What are the appropriate benchmarks
for comparing these machines?
- Impacts of the requirements
of applications areas like climate
research, nanotechnology, atomic energy, and biotechnology on the
future of high
end
computing.
- Current and planned
applications of the Earth Simulator. The
Earth
Simulator was originally announced as a tool for research in Earth
sciences, and
is now being used for that purpose. It is especially suited for
simulating
complex linked systems, for example, linked global atmosphere and ocean
models.
Recent government programs, however, have gone beyond the application
in Earth sciences. How would ES open up and possibly change the
research outlook
for other computational sciences, such as biology, genomics, medicine,
nanotechnology, or even human or artificial intelligence? How are the
Japanese industries making the best use of the ES for research and
commercial applications? What has been the payoff so far in terms of
science and engineering advancements enabled by the Earth Simulator?
- Experiences of the users -
do they support investment in more
powerful parallel vector supercomputers? What have been the experiences
of U.S.
or other foreign scientists in using the Earth Simulator?
Non-technical Issues
- R&D funding for the
Earth Simulator project and other
Japanese supercomputer projects in progress (e.g., the Petaflop Protein
Explorer launched last
year and the Fujitsu machine for NAL) in general, compared to the
United States.
How strong is Japanese government support for the Earth Simulator and
other
related programs?
- Overall government strategy
for supercomputing R&D and
applications development - does Japan have an industrial policy for
developing and maintaining its domestic supercomputing hardware or
software development capabilities, and if so, what is it? Is there a
Japanese road map for supercomputer development?
- Opportunities for
collaboration between the U.S. and Japan.
- Management of the Earth
Simulator's development and other high
performance
computing projects, especially the interaction among government,
industry, and
academe. How do interactions and collaborations among these groups
involved in
the Earth Simulator project compare to other major Japanese government
projects
involving shared funding and use of such resources?
- Implications of the Earth
Simulator and other advanced Japanese
machines in
future supercomputing R&D, manufacture, and world markets.
- Roles of industry and
academia in research, technology
development,
and commercialization of high end computing.
- Long-range research,
educational, and infrastructure issues that
need
to be addressed to promote better U.S. progress in the field, and to
regain
U.S. world leadership.
![[photo: Al Trivelpiece]](pics/trivelpiece_a.jpg)
Al Trivelpiece
(panel chair) |
- Former director of the
Oak Ridge National Laboratory
- 14 Wade hampton Trail
Henderson, NV 89052
- (702) 492 1602
- awt511 at cox.net
|

Rupak Biswas |
- NASA Ames Research
Center
Mail Stop T27A-1
Moffett Field, CA 94035
- E-mail:
rupak.biswas at nasa.gov
- Office: 650-604-4411
- Fax: 650-604-3957
|

Peter Paul |
- Deputy Director
for Science and Technology, Brookhaven
National Laboratory
- Brookhaven National
Laboratory
Upton, NY 11973-5000
- 631-344-8623
- ppaul at bnl.gov
|
![[photo: Jack Dongarra]](pics/dongarra_j.jpg)
Jack
Dongarra |
|

Katherine
Yelick |
- Professor
- 777 Soda Hall
Computer Science Division
University of California at Berkeley
Berkeley, CA 94720-1776
- Phone: (510)642-8900
- Fax: (510)642-5775
- yelick at
cs.berkeley.edu
|
Advisor to the Panel