Using MuMMI to Model and Optimize Energy and Performance of HPC Applications on Power-Aware Supercomputers
Authors: Xingfu Wu (Texas A&M University), Valerie Taylor (Texas A&M University)
Abstract: The MuMMI (Multiple Metrics Modeling Infrastructure) facilitates systematic measurement, modeling, and prediction of performance and power consumption, and performance-power tradeoffs and optimization for Power-aware HPC systems. In this paper, we use the MuMMI to model performance and power. These models focus on four metrics: runtime, system power, CPU power and memory power. We rank the counters from these models to identify the most important counters for application optimization focus, then demonstrate the counter-guided optimizations with an aerospace application PMLB executed on two power-aware supercomputers, Mira at Argonne National Laboratory), and SystemG at Virginia Tech.
Two-page extended abstract: pdf