lecture image CCT Colloquium Series
Integrated Performance Monitoring: HPC Workload Characterization
David Skinner, Lawrence Berkeley Lab
Johnston Hall 338
March 16, 2007 - 03:00 pm
This talk introduces the Integrated Performance Monitoring (IPM) framework which is in use at several HSF and DOE centers as a means of understanding HPC application performance in production computing settings. The creation of a portable, scalable, and low overhead profiling infrastructure for HPC applications is driven by the need to both optimize HPC applications on a given architecture and to characterize the performance of applications across architectures. The latter allows for making optimal matches of HPC workload and HPC architecture.
Speaker's Bio:
David Skinner is currently a group leader in high performance computing at Lawrence Berkeley Lab. His Open Software and Programming group solves problems related to shaping HPC resources into high performance tools for scientific research. Work in his group focuses on performance analysis, information management, and the deployment of computing and data services at the NERSC center. He received a PhD from U.C. Berkeley in 1999 in chemistry for work applying parallel computing to quantum and semiclassical chemical reaction dynamics. He received a B.S. chemistry from Kansas University in 1995. His research interests include quantum reactive scattering, molecular dynamics of large systems, and architectural analysis of computing systems.
This lecture has a reception.