Performance Monitoring, Prediction, and Run-Time Adaption of Grid-Distributed Applications

78647 Three-dimensional seismic imaging is a computationally-intensive technology that has allowed the success ratio of exploratory oil and gas wells to increase from 20% to 80%. This project will develop technology for identifying, predicting, and monitoring factors to increase the efficiency of this process. Specifically, given a parallel program executing on a homogeneous or heterogeneous cluster, with certain array sizes, memory allocations, and data flow characteristics, the objective is to find the set of application-specific parameters and code implementations that minimize runtime. Phase I will identify key parameters that influence parallel runtime and study the interactions between these parameters; use monitoring tools to study the performance characteristics of target applications on thousands of processors; develop a ¿planner¿ that, in conjunction with a dynamic optimization program, searches for a set of parameters for which the analytic model performs optimally; integrate the monitoring and dynamic performance optimization tools into the infrastructure for the Internet Seismic Processing System; and research the runtime optimization of very computing intensive geophysical applications. Commercial Applications and Other Benefits as described by the awardee: The net benefit to the global oil industry from three-dimensional seismic imaging, through reduced drilling costs and additional reserves exploited, amounts to $11 billion a year. The new seismic imaging technology should increase effectiveness in characterizing oil reservoirs and monitoring enhanced oil recovery processes, thereby opening up new exploration venues in extremely complicated geological conditions like the deep Gulf of Mexico area. Future applications should include facilitating computations related to pharmaceuticals, DNA, aerospace, finance, and entertainment.