Computing at Scale
Wednesday morning starts off looking at one of the most talked about visions of the HPC future to come on the scene in the recent past: Parallel Computing Landscape: A View from Berkeley. David Patterson, one of the report’s principal authors, discusses A View in this invited talk.
The Programming Models papers session looks at the role of MPI in application development time, an adaptive cut-off for task parallel frameworks, and studies the software environment on the Intel 80-core terascale processor.
Finally, the Doctoral Research Showcase highlights the work of Chao Wang in developing a mechanism for process-level fault tolerance for job healing in HPC, a topic relevant not only to tomorrow’s very large scale computers, but also to the creation of a more robust computational support infrastructure.
Wednesday’s activities related to the computational infrastructure theme start off with a panel discussion. The provocatively named Will Electric Utilities Give Away Supercomputers with the Purchase of a Power Contract? panel explores the crunch datacenter budgets are feeling right now as falling computer prices push their power distribution systems to their limits.
Chao Wang’s paper on process-level fault tolerance, discussed above, also points to key new technologies of interest for future computational infrastructure developments.
Wednesday’s application horizons-themed activities focus on opportunities for HPC in medicine and biology. First up is invited speaker Kenneth H. Buetow from the National Cancer Institute with discussion of the issues involved in developing a framework that will enable personalized medicine. A framework in this application area will involve bringing together many different communities and standards of practice, topics closely related to the expanded access theme as well.
Later in the day, SC08 presents two Masterworks sessions related to HPC and the biosciences. HPC in the “Personalization” of Cancer Therapy: Genomics, Proteomics, and Bioinformatics examines the technologies needed to tailor therapy to the molecular profile of an individual patient’s disease. Computational Opportunities in Genomic Medicine describes some of the challenging computational problems in basic biology and medicine, and outlines the software infrastructure that is needed to support this highly interdisciplinary field.