Performance and Power Characterization of HPC Applications on Heterogeneous Memory System

PI Name Zhiling Lan
PI Institution University of Illinois Chicago
Collaborating ANL Division Mathematics and Computer Science (MCS)
Project Description

The goal of this project is to characterize performance and power consumption of various HPC applications, including conventional scientific simulations and emerging ML applications, on heterogeneous memory system composed of DRAM, persistent memory (e.g, Intel Optane DC), NVMe, and/or disaggregated memory. Specifically, we will conduct experiments with Seven Dwarfs, NPB, and Candle benchmarks. Our experiments will be focused on characterizing runtime and power consumption of executing different applications under different memory configurations. This project is a collaborative work with Valerie Taylor & Xingfu Wu at MCS and Bill Allcock at ALCF.

Testbed

Skylake or other testbed with Intel Optane DC