Session: The Climate/Ocean/Weather and HPC Communities: The State of the Union
  Chair: Jay Larson (Argonne National Laboratory), John Drake (Oak Ridge National Laboratory)
  Time: Wednesday, November 19, 5:00PM - 6:00PM
  Rm #: 42-43
  Jay Larson (Argonne National Laboratory) and John Drake (Oak Ridge National Laboratory)
  Climate/Ocean/Weather applications constitute a grand challenge in high-performance computing. During the 1990's, this community -- like other computational science communities -- worked hard to migrate its software from vector platforms to commodity-based multiprocessors. This effort proved long and arduous, and some found it difficult to attain satisfactory levels of efficiency (defined as a fraction of peak performance) on these platforms. At Supercomputing 2001, one bird-of-feather session asked the question: Is US climate modeling in trouble? This question stemmed from discontent among US climate scientists over the 1997 Department of Commerce anti-dumping order that effectively blocked the import of Japanese vector supercomputers, and the impression that this ban gave their foreign counterparts a substantial competitive advantage. Since the 2001 BoF, much has changed in terms of hardware and software technology. Vector platrforms have made a comeback in the form of the Earth Simulator and the Cray X-1 (an issue that is the topic of a separate BoF). Efforts to enhance application performance portability on commodity-based multiprocessors are yielding promising results. Grid computing research and development continues to advance. A push to explore new hardware architectures capable of achieving petaflops-scale performance will pose new performance-portability challenges. Component-based software engineering is emerging as a catalyst for scientific library interface standardization and elimination of language barriers.

These circumstances give impetus to a broad set of questions: What do climate/ocean/weather scientists require and how can the HPC community help? What is the present and future impact of large software engineering efforts such as SciDAC/CCSM project, the Common Component Architecture Forum and NASA's Earth System Modeling Framework? How well will current efforts meet the challenges posed by new processor/platform architectures? What is the current and future role of grid computing in climate? This bird-of-a-feather session will be a forum in which we attempt to answer these questions.
  Link: --