CCS User Exchange Forum
The UM modeling community has started to take advantage of the
hardware/software/personnel resources that the HPC initiative has recently
put in place under the umbrella of the Center for Computational
Science. The computing/networking and storage capacity are currently
(as of 8/26/2008) adequate enough to permit serious (publication-grade)
simulation to take place.
The hardware/software/personal expansion alone is not enough to fulfill
the vision of the Computational Science Initiative at UM. This expansion
must be matched by the growth of a user community to anchor the center's
scientific aims, provide its administrators with the user's requirements,
and spread the (computational) knowledge base among the center's users
and the university community at large. To this date the user community
remains dispersed among the 3 university campuses, often confined to
semi-enclosed research group.
The aim of this series of informal talks is to foster the
formation of high performance computing user community within
UM. In summary this community building effort aims to:
-
Increase awareness of Computational Science among UM students and researchers.
-
Enhance interdisciplinary interaction among faculty,
research staff and students with a view to encourage collaboration
at a unviersity-wide level.
-
inform CCS' administrators about their users'
computational/training requirements, and guide
CCS's users training and future planning.
-
Potential for course development that can serve
interdisciplinary
education and research purposes at RSMAS.
-
nurturing a modern, high-tech, intellectual flavor for
research at
RSMAS.
The following requirements must be considered in starting to put this effort together:
Event logistics:
University-wide communication: The user community is currently spread
on 3 campuses, it is imperative to bring them together if we are to
build university-wide links. We suggest thus that the venue for the
meeting/presentation be rotated among the campuses. An opening ceremony
at RSMAS would be useful as it can naturally evolve into a happy hour on
a Friday afternoon. The frequency of the meetings is an issue also as
most people are already to committed to their departmental seminars. If
weekly presentation prove to be too much, then a monthly schedule maybe
be more appropriate.
Event format
The effort will be hampered by the different computing cultures of the
different users (CFD versus genomics for example) and the different
expertise levels (HPC developers versus users of Black Box models). The
most effective route to foster collaboration is to discuss the nuts and
bolts of HPC in terms of computing needs, algorithms, programming, etc...
It then becomes necessary to limit the temptation to razzle dazzle
with PowerPoint presentation, and focus on algorithms. It would be then
useful to require that most presentation be done on a white board with
no access to PowerPoint or slides (discussions on these issues would be
useful). This would tend to favor algorithm discussion which may or may
not be appropriate to the UM community.
Event Administration
The effort must be lead by the user community with support from the
CCS center. The onus will then be on the user community to organize and
coordinate this effort. Reliance on the CCS may, first overburden the
latter at a time when it is still busy setting up the system, and two
re-enforce a sense of reliance/dependency on the center's staff.
Nevertheless, the effort must be coordinated with the CCS staff as they
are so far the only one with a bird's eye view of the user community.