Daniel Reed - International Scaling: Realizing the Potential of Grid and High-End Computing

Abstract

Large-scale grids containing thousands of sites are being considered, developed, and deployed. Similarly, node counts for terascale systems have grown to tens of thousands, with petascale system likely to contain hundreds of thousands of nodes. In addition, a tsunami of new experimental and computational data poses equally vexing problems in analysis, transport, visualization, and collaboration. We must rethink traditional assumptions about software scaling, component integration, and hardware reliability. Our thesis is that the "two worlds" of software-grids and parallel systems-must meet, embodying ideas from each, if we are to build a usable and useful infrastructure. This talk describes approaches to scalable performance measurement, dynamic adaptation, and grid integration and their implications for large-scale science and engineering.