Does Varying BeeGFS Configuration Affect the I/O Performance of HPC Workloads?

Arnav Borkar, Joel Tony, Hari Vamsi K. N, Tushar Barman, Yash Bhisikar, Sreenath T. M., Arnab K. Paul
October, 2023
Cite DOI URL

Abstract

The increasing gap between processor and disk performance leads to high performance computing (HPC) applications facing I/O bottlenecks. This makes parallel file systems one of the most important components in an HPC cluster. This work analyzes the I/O performance of different workloads for various BeeGFS configurations. Our analysis shows that the default allocation strategies and striping configuration leads to an imbalanced distribution of workload data, thereby negatively affecting the I/O performance.

Type
Conference paper
Publication
2023 IEEE International Conference on Cluster Computing Workshops (CLUSTER Workshops)
Adaptive Systems File Systems Conferences Cluster Computing Resource Management BeeGFS High Performance Computing Parallel File System
Joel Tony
Joel Tony
Computer Science Undergraduate

My research interests include distributed systems, storage systems, high-performance computing and MLSys.