The rapid developments leading to supercomputers with hundreds of thousands to a million processors, now being funded by NSF and DOE, brings with it opportunities to study turbulent flows with increasing fidelity in a number of geophysical and astrophysical fluid dynamic (GAFD) settings. Such "petascale initiatives" in supercomputing, promising to yield sustained petaflop performance on realistic problems, also raise many challenges, since few current scientific codes are likely to scale efficiently to such large number of processors or cores. We are using this mini-symposium to bring together some colleagues to discuss the challenges in utilizing petascale resources on GAFD problems dealing with turbulence and dynamos in round settings, whether in the solar convection zone or in the geodynamo. Many of the same challenges are likely to be shared with global atmosphere and ocean models, and so too with climate models, since the need to solve an inherent Poisson problem leads to intense communication demands between the processors.
In order to take advantage of the promised new machines, as in the recent NSF Tier 1 award announcement, do we need new programming paradigms to write codes that will be able to use these systems efficiently? Are these paradigms mature enough yet? In reviewing some of the numerical methods currently in use in GAFD, we will seek to assess whether there are inherent limitations in using these methods on petascale systems. Do some of these methods have inherent strengths regarding scalability, performance, or time-to-solution on these systems? What tools must be organized to deal with the analysis of the vast data sets that will result from the simulations? A further challenge to new code development involves uncertainties about the actual machines that will be built, since at this stage many proposed features of the petascale architectures have yet to be realized and tested. However, TeraGrid access to some of the new Tier 2 machines will provide the means to test both modified existing codes and new codes as we await the arrival of the sustained petaflop machines.
This Turbulence and Dynamos at Petaspeed meeting seeks to address some of these questions by bringing together some practitioners in scientific computing in a collegial atmosphere in which we can share ideas and experiences on these complex issues. In addition to addressing some of the challenges in utilizing petascale resources, we hope this meeting will foster collaborations and longer-term discussions between attendees and institutions. We invite all who may be interested in coming to the presentations and discussions, since all events will be in open settings.
This meeting is hosted jointly by the Institute for Math Applied to Geosciences (IMAGe) at NCAR and by the Joint Institute for Laboratory Astrophysics at the University of Colorado. It serves as a follow-on to a series of informal meetings that occurred in Spring 2007 between these two institutions to discuss new computational approaches to deal with GAFD simulations in the impending petascale era. It was recognized that the new technologies are likely to have a tremendous impact on many areas of scientific research, but that there are also significant obstacles to overcome in using the petascale systems efficiently. Some funding for this meeting comes from the NSF cooperative agreement with UCAR and from the NASA Heliophysics Theory Program grant to JILA.
Annick Pouquet, NCAR, Institute for Mathematics Applied
Duane Rosenberg, NCAR, Institute for Mathematics Applied
email: duaner (at) ucar dot edu
Phone: 303 497 1636
Juri Toomre, University of Colorado, JILA
email: jtoomre (at) solarz dot colorado dot edu
phone: 303 492 7854
John Clyne, NCAR, Data Analysis Services Group
John Dennis, NCAR, Computer Science Section
Paul Fischer, Argonne National Laboratory
Akira Kageyama, The Earth Simulator Center, Japan Agency for Marine-Earth Science & Technology
James Lottes, Argonne National Laboratory
Dimitri Mavriplis, University of Wyoming
Mark Miesch, NCAR, High Altitude Observatory
Ramachandran Nair, NCAR, Institute for Mathematics Applied to Geosciences
Alan Norton, NCAR, Data Analysis Services Group
Annick Pouquet, NCAR, Institute for Mathematics Applied to Geosciences
Piotr Smolarkiewicz, NCAR, Institute for Mathematics Applied to Geosciences and Mesoscale & Microscale Meteorology
Amik St-Cyr, NCAR, Institute for Mathematics Applied to Geosciences
Mariana Vertenstein, NCAR, Climate and Global Dynamics Division
Michael Wiltberger, NCAR, High Altitude Observatory
- Katherine Yelick, University of
Meetings will be held at the National Center for Atmospheric
Research's Center Green Campus in Boulder, Colorado, 3080 Center Green Drive, 303 497 2525.
Click here for directions
JILA is one of the nation's leading research institutes in the physical sciences. A joint institute of the University of Colorado and the National Institute of Standards and Technology, JILA is located on the CU campus. JILA's NIST members hold adjoint faculty appointments at CU.
The Institute for Mathematics Applied to Geosciences is a group embedded
within the National Center for Atmospheric Research
(NCAR) for the purpose of
advancing mathematical theory and its application to all facets
of NCAR and the geophysical community at large. IMAGe is substantially
funded by the
National Science Foundation.
NCAR was formed in 1960 and has a broad interdisciplinary research
program involving more that 1000 employees of which several hundred
hold advanced scientific or engineering degrees. The NCAR scientific
program includes nearly all aspects of the atmosphere including
climate and weather, atmospheric chemistry, ecology, instrumentation,
scientific computing, and economic and societal impacts of
JILA scientists explore some of today's most challenging and fundamental scientific questions about quantum physics, the design of precision optics and atom lasers, the fundamental nature of matter, and processes that shape the stars and galaxies. JILA science encompasses seven categories: astrophysics, atomic & molecular physics, biophysics, chemical physics, nanoscience, optical physics, and precision measurement.