Why the June 1995 ERSUG Meeting is Important
Bill McCurdy describes below a competitive process through which a
decision will be made by MICS Division (formerly the OSC) in the June, 1995
(1) possibly move NERSC to Lawrence Berkeley National Laboratory, and
(2) redefine to some extent the mission of the Center.
All of this would be effected within a significantly reduced cost envelope.
The LLNL proposal to keep NERSC where it is and the LBL proposal may soon
be available for review and comment.
This meeting represents an excellent forum for users to make their
views known. If you depend on the service offered by this Center,
we strongly urge you or a representative from your institution to
attend this meeting.
If you wish to attend the ERSUG meeting, please find details on the
nersc.meetings.workshops Bulletin Board or by accessing the
ERSUG meeting information in the "What's New" section of the
NERSC home page (http://www.nersc.gov).
The following is a message from Bill McCurdy to the Executive Committee
of the Energy Research Supercomputer User Group (ExERSUG). ExERSUG members
have asked that it be posted for the NERSC community. If you have questions
or comments, or want to share your views on this issue you can post your
views to the nersc.general Bulletin Board from the Cray-2s or SAS.
example: Pnews nersc.general
You may also wish to e-mail comments to your program officer.
Dear ExERSUG members,
This is a brief note to give you some information about the process at
DOE Headquarters which is affecting the future of the NERSC.
Here are the facts:
As most of you have heard, the Mathematical, Information and
Computational Sciences Division (MICS, formerly the Office of Scientific
Computation) has asked for a proposal from the Lawrence Livermore National
Laboratory and, as I understand it, from Lawrence Berkeley Laboratory for the
operation of the NERSC. The program they are reviewing is a combination of
three components: (a) the High Performance Computing Access Center (Access
Program), (b) the Energy Sciences Network (ESnet), and (c) associated
computational science constituted primarily by the ER-funded Center for
Computational Science and Engineering at Livermore. The move to reassess the
program is entirely a Headquarters decision.
We have been asked to address a 15% to 20% reduction in funding
overall, and have been given a list of criteria against which the proposals
will be evaluated which focus on the combination of cost savings and benefits
to Energy Research programs. A review panel will be convened to evaluate the
proposals. To reach headquarters on time, our proposal has to be sent out this
Friday, May 12.
I and the NERSC staff have prepared an aggressive and extensive
proposal which meets the cost reduction goals of ER, but we are determined to
maintain a high level of service at NERSC within that constrained budget.
We have protected the acquisition of new hardware in the budget, so that you
will have the best we can offer in computational facilities, and consequently
we have streamlined every other aspect of the operation of the Center to meet
the budget targets.
Since we are in a competitive situation, I cannot discuss many of the
details of the choices we have made. Those discussions, which I believe I
and the NERSC staff owe to you as users who depend on the Center, will be
possible in a few weeks. I regret that, because of the competitive process
and short time we were given to respond, we could not involve ExERSUG and
the other users in the proposal preparation directly. I cannot bear it when
someone asks me to just trust them with my scientific research resources,
and I could not bear to ask that of you. I and the NERSC staff have only
our track records with you to stand on.
I can say that we have incorporated, in an aggressive manner, the
advice we were given about cost effective hardware by ERSUG and ExERSUG
at the last meeting. Those discussions were fresh in our minds as we
wrote the proposal. This proposal process gives us the chance to implement
some of that advice on a shorter time scale than might otherwise have been
possible. At this point that is all I am comfortable saying.
An issue for the users and for EXERSUG:
NERSC is not broken, but it can be made better, and I believe it can be
made better while spending less money doing so. By the ERSUG meeting we will
be able to discuss all aspects of our proposal for NERSC's future. Between
now and then I think that it is important for ER program officers to hear how
important the computational resource and service aspects of the program are
to their researchers. NERSC must maintain a correct balance between service
(both computing and networking) and technology development and infusion into
the computing environment. Part of our service is in collaboration with users,
as you know, including parallelizing codes and helping researchers take
advantage of the Special Parallel Processing program. The service components
of NERSC are like the power provided by a public utility, nobody complains and
few comment until the power goes off. I think that it is an issue for the
users that the importance of the service functions not be eclipsed by the high
gloss of new technology development, which does however have its place.
The users of NERSC have the opportunity to make their opinions about
the program and its future known to ER program officers. Besides your own
program officers there are also the members of the Supercomputer Access
Committee who make allocations at headquarters and whose names you can get from
Tom Kitchens in the MICS Division.
There is of course much that we could discuss about the future of NERSC
and how it will be charted during the decision process at headquarters. I am
under constraints imposed by the competitive process which will relax in the
coming weeks. I look forward to discussions with you, and I am doing my
absolute best to act in the interest of the users at all times.