FYI: 15 supercomputers to be purchased by GOJ by 31 March 1994.

[Dave Farber <farber () central cis upenn edu>]

From: 
 Dr. David K. Kahaner
 US Office of Naval Research Asia
 (From outside US):  23-17, 7-chome, Roppongi, Minato-ku, Tokyo 106 Japan
 (From within  US):  Unit 45002, APO AP 96337-0007
  Tel: +81 3 3401-8924, Fax: +81 3 3403-9670
  Email: kahaner () cs titech ac jp
Re: Supercomputer sales to Japanese government facilities
26 May 1993
This file is named "sc-new.93"

ABSTRACT. 1 April 1993 -- 31 March 1994 planned supercomputer sales to
Japanese government facilities are given.

Interest in Japanese supercomputer and networking activities is very
high, not only here in Japan, but also in the West. Over the past month
or so, a variety of news reports have been circulating. In a few cases I
have remarked that some of these reports were inaccurate or only
partially accurate.  Recently I met with a key MITI official in order to
get as much up to date information as possible.

      Mr Hidetaka Fukuda, Deputy Director
      Industrial Electronics Division
      Machinery and Information Industries Bureau
      MITI
      1-2-1 Kasumigaseki, Tokyo 100 JAPAN
       Tel: +81 3-3501-1074; Fax: +81 3-3580-6073

As part of a regular MITI rotation cycle, Fukuda will be moving to
another position soon. Nevertheless, he still retains a direct interest
in supercomputer activities, and speaks for the Ministry.

Mr Fukuda explained that the Japanese government had planned to purchase
four supercomputers as part of a regular (ordinary) budget process.

     Ministry of Education:

          Tohoku University Computer Center
          
          Okazaki Institute for Molecular Science

          The Institute of Space and Astronautical Science

     Meterological Agency:

          Meterological Research Institute

In addition, as part of the supplemental budget, which was, in part,
approved as an economic stimulus for the country, 11 additional
supercomputers will be  purchased (for a total of 15) during
the current fiscal year, 1 April 1993 -- 31 March 1994.

     Ministry of International Trade and Industry:  

          Agency of Industrial Science and Technology, 
              Research Information Processing System

          Angstrom Technology Partnership (Private)(2)

     Ministry of Education:

          The Research Institute for Iron, Steel, and Other
              Metals, Tohoku University

          University of Tsukuba

     Science and Technology Agency:

          National Aerospace Laboratory

          The Institute of Physical and Chemical Research (RIKEN)

          Power Reactor and Nuclear Fuel Development Corp.

     Ministry of Posts and Telecommunications:

          Communications Research Laboratory

     Ministry of Health and Welfare:

          National Cancer Research Center Institute

     Ministry of Agriculture, Forestery and Fisheries:

          Institute for Agrobiological Resources

The term "supercomputer" used here is generic, and includes vector
supercomputer as well as highly parallel computer.

Mr Fukuda stated that the procurement process will be fully competitive,
as per the GATT agreement and established procedures to introduce
supercomputers.  I commented that within the Japan science community it
was commonly believed that a 50/50 split between Western and Japanese
vendors was expected, but he rejected this idea as completely false. (It
is true, however, that key people at some sites have strong preferences
toward particular vendors based on installed systems, personal contacts,
etc.)  Thus, in principle, all the systems purchased could be Western
(or Japanese), and Fukuda agreed this was possible.  If all were
Western, that would not appear to me much of an economic stimulus to
Japan, but Mr Fukuda claimed it did not matter because the fraction of
the total stimulus package associated with this item is modest.

Of some concern in the West is in the method of evaluation of
supercomputer proposals in Japan, and both Fukuda and I agreed that,
although agreements are in place to quantify and make the process more
objective, there are still some aspects that are not easily
quantifiable. I felt that supercomputer vendors were usually in a good
position to compare relative performance characteristics, but suggested
that better access to Japanese supercomputers by Western scientists for
benchmarking purposes might make the evaluation more objective. He
agreed with this and said that he would encourage and might even support
such proposals; his thinking is that this could actually result in
Western organizations selecting Japanese supercomputers for their
procurements.

Mr Fukuda described, briefly, the need for Japan to improve its
networking capacity. He also stated that it was generally believed (in
Japan) that the US HPCC initiative was a response to earlier activities
in Japan, and that now Japan was thinking about how to respond to that.
This statement was astonishing to me, as I believe that the HPCC efforts
in the US have been developing naturally and Japanese activities were
not serious considerations during its formulation.

I also asked for Mr Fukuda's reaction to various reports in the US press
about a Japanese project known as Mandara (or Mandala), which purports
to be an effort to develop a large number of Japanese supercomputer
centers and link them together with a gigabit network. While not
fundamentally unreasonable, the English language unofficial translations
about Mandala are written in strident terms vis a vis US-Japan
competition, and include concerns that the HPCC activities in the US
will have a serious detrimental impact on Japan, resulting in a "long
term decrease in GNP and the standard of living" in Japan.  Mr Fukuda
stated, emphatically, that he (and hence MITI) was totally unaware of
such a project, and assured me that it is not a Japanese government
plan, either in existence now or soon to be put into place! He also
stated that it is not MITI's policy to respond to Western reports,
exceptions occurring sometimes if important sources (NY Times, etc) are
involved.  Nevertheless, translations of the report state that MITI will
obtain financing for the plan by spending 100B Yen per year for ten
years.  Clearly, somewhere, there is a break in the communications loop.
After checking, Fukuda wrote to tell me that Mandala "seems to be
promoted by some organizations related to STA [Science and Technology
Agency, part of the Prime Minister's Office], but MITI or all Government
of Japan doesn't authorize or know clearly about it."

Finally, I asked about the new Asia Pacific Super Computing Design
Project", which was recently a topic of discussion between Prime
Ministers of Japan and Australia. Mr Fukuda explained that Australia, a
country with a weak manufacturing base, wants to collaborate with
Japanese vendors in supercomputer development.  (Australia is already an
strong research partner to Fujitsu in activities related to its AP1000
parallel computer.) Fukuda also commented that at the current time,
there are discussions with several other Asian countries on this topic.

----------------------------END OF REPORT--------------------------------

------ End of Forwarded Message