cunningh@noscvax.UUCP (Robert P. Cunningham) (12/25/84)
[abstracted from Datamation, Dec 1 issue]
Some 22 groups have applied for a piece of the National Science
Foundation's (NSF) $18 million worth of initial funding in their
Advanced Scientific Computing Centers program. Most are consortiums
of different universities around the U.S.
According to Datamation, the 22 proposals have been trimmed to six or eight,
and will be trimmed to three in February. The winning supercomputer
centers are scheduled to begin operations on 1 July 1985, and continue
under NSF funding for five years. The NSF hopes to open 10 centers by
1990.
Although the identities of the leading groups are officially secret,
Datamation mentions as favorites:
the Consortium of Scientific Computing, consisting of 12
mostly-eastern universities.
the University of Illinois.
the West Consortium lead by General Atomic Inc., and including U.
of Wisconsin, U.C. San Diego, UCLA, Stanford, CalTech, U. of Hawaii
and others.
a combination of the National Center for Atmospheric Research (NCAR)
and the University Corporation for Atmospheric Research.
And a second tier of contenders including:
Cornell University.
a Washington consortium consisting of Boeing, U. of Washington and
George Washington University.
a Houston Area Research Consortium.
U. of Minnesota.
Purdue University.
U. of Michigan.
a consortium of Carnegie-Mellon University & Westinghouse.
The purpose of NSF's program is to significantly increase access to
supercomputers by researchers, as outlined by the 1982 study of the Lax
Panel on Large Scale Computing in Science and Engineering sponsored by NSF
and DoD. The panel found that: important segments of the research &
defense communities lack effective access to supercomputers; students are
neither familiar with their capabilities nor trained in their use; access
to supercomputers is inadequate in all disciplines; the capacity of today's
supercomputers is orders of magnitude too small for currently urgent
problems; the current development of those computers will yield only a
small fraction of the capability technically achievable; computer
manufacturers have neither the financial resources nor commercial
motivation to undertake the requisite exploratory r & d.
The number of supercomputer installations in the U.S. are less than in
Europe or Japan. Most are at -- and only available to -- government labs.
The only American universities with them at present are Minnesota, Purdue
and Colorado State. Mainly because they're very expensive. A Cray goes
for anywhere from $4-14 million + $1.5-5 million for peripherals + site
preparation + approx. $30k/month for the service contract.
NSF, with a Congressional FY85 apropriation of $40 million in hand, now
aims to provide the research community with supercomputer
centers of excellence that will become major nodes on a future national
network.
Datamation notes that the main competition may actually be between Cray and
CDC. Ten proposals involve Crays ranging from the smallest X-MP to the
high-end Cray 2. Seven applicants requested Cybers. Other machines
suggested by potential participants include FPS boxes driven by IBM front
ends, three Denelcors and one Amdahl.
Datamation's opinion: two awards will be given to proposals involving
Cray machines and one to a proposal involving a CDC Cyber 205 or
higher-level machine such as ETA's GF-10.
--
Bob Cunningham {dual|ihnp4|vortex}!islenet!bob
Honolulu, Hawaiigrunwald@uiucdcsb.UUCP (12/31/84)
In regards to the super-computing rumour and UIUC.
UIUC is definitly getting funding for a Cray-2 or X-MP (not sure which) which
will be managed by Larry Smarr from the Physics department here. They are also
acquiring about 150 (rumoured) Sun workstations for front/back-ending to the
Cray & for data display, etc.
Someone told me that the person who runs Cray now is a UIUC alumni,
so we're going to get a deal similar to the deal that Seymor Cray gave to
U of Minn.
Additionally, two new labs are in the process of being set up, also with NSF
funding (from my info). The Center for Scientific and Engineering Super
Computing (CSESC) and another lab whose name I've forgotten.
CSESC will continue the work of Kuck et al on the Paraphrase system
for concurrency explotation in programs & work into general concurrency &
parallel systems.
The other lab is going to design & build the Cedar architecture
machine using processors & a shuffle-exchange network between them and
heirarchical memories.
This two labs are public knowledge, and hiring has been going on for
some time (see IEEE Computer magizine, back pages). The Cray grant has been
written up in our local newspaper. The Sun stuff is rumour, although it was
offered as the reason why Bill Joy made a stop out here last spring.
Purdue already has a Cyber 205. From my understanding, they need
extra NSF money to maintain the machine, not to buy new ones.
dirk grunwald
university of illinois, urbana-champaign
grunwald@uiuc.arpa
grunwald%uiuc@csnet-relay
{pur-ee, ihnp4}! uiucdcs ! grunwald