Practical PDES: Using PDES to Better Understand HPC Systems
Large-scale high-performance computing systems employ cutting edge hardware and software to serve the needs of complex and highly parallel computational science applications. Especially when new technologies become available, facilities procuring these systems need assistance in assessing the relative merits of different architectural features, and need to plan how to best take advantage of these systems in advance of their deployment. Parallel discrete event simulation could be widely used in these planning processes, but so far its impact has been limited. Using a current question of interest to facilities, we will explore the requirements for employing PDES towards understanding modern HPC platforms, discuss some efforts in these directions, and point towards opportunities for new research and development to make PDES a central tool for HPC system planning.
Robert Ross is a Senior Computer Scientist at Argonne National Laboratory and the Director of the DOE SciDAC RAPIDS Institute for Computer Science and Data. Rob’s research interests are in system software for high performance computing systems, in particular distributed storage systems and libraries for I/O and message passing. Rob received his Ph.D. in Computer Engineering from Clemson University in 2000. Rob was a recipient of the 2004 Presidential Early Career Award for Scientists and Engineers.