Traces of parallel programs contain an overwhelming amount of data, often making it challenging to identify the root causes of performance issues. This talk presents a methodology that synthesizes trace data into a Directly-Follows Graph, enabling comparative analysis of program behavior. This methodology is implemented in the STrace Inspector tool, which synthesizes large amounts of system call traces for benchmarking and comparison of I/O activities in HPC applications.
Reference Material
About the Presenter
Aravind Sankaran has been a scientific researcher at the Jülich Supercomputing Centre (JSC) since 2023. He works in the Algorithms, Tools, and Methods Lab responsible for application optimization and user service tools. He holds a Master’s degree in Simulation Science and has been pursuing a PhD on topics related to application benchmarking and performance comparisons since 2020 at RWTH Aachen University.
He is currently involved in developing tools to investigate jobs running on JSC clusters, as well as providing user support and training.