Akshar Patel

Advisor: Zhiling Lan

Undergraduate: University of Alabama (Computer Science)

Graduate: University of Illinois Chicago (Computer Science)

Project: Resource Usage Profiling for High Energy Physics Applications on HPC Systems

This project investigates methods for profiling high energy physics (HEP) applications based on their resource utilization patterns on high performance computing systems. The work focuses on analyzing power and memory usage data generated by large-scale HEP workloads in order to better understand application behavior during execution. The study uses the Fermilab SPINE application running on the Polaris system at Argonne National Laboratory as a primary case study. By examining time series resource usage data, the project evaluates profiling techniques that can capture distinctive patterns in application execution, including approaches such as chaos game representation and other time series analysis methods. The objective is to determine whether resource utilization patterns can be used to generate application specific signatures or "fingerprints" for HEP workloads. Such fingerprints could support improved job scheduling, workload classification, and resource management strategies for large scale computing systems.