Accelerating insights gleaned from leadership-class supercomputing simulations

November 1, 2011

Argonne computer scientists have devised a software-based approach, called GLEAN, to improve I/O performance on its IBM Blue Gene/P by exploiting the machine's network topology to intelligently aggregate data. The approach provides a way to perform real-time data analysis and visualization during large-scale simulation runs, enabling scientists to obtain insight more quickly. The approach has been demonstrated using synthetic and application benchmarks to yield multifold improvement on the BG/P. Software-driven topology-aware approaches will be especially useful as the complexity of supercomputing systems continues to increase. More ยป