Advancements in High Performance Computing Cluster Resource Utilization through a Comprehensive Monitoring Dashboard

This paper describes developing and deploying a browser-based High-Performance Computing (HPC) monitoring dashboard to track HPC jobs performance metrics for both CPU and GPU workloads on the ANTYA HPC cluster at the Institute for Plasma Research (IPR). Considering the importance of cluster resources, the authors have enabled tracking of job performance and job failures in using resources more efficiently to maximize resulting scientific knowledge. The dashboard uses historical and real-time data generated continuously from the scheduler installed in ANTYA. The front-end visualizations generated from the raw data will help the system administrator to quickly view and analyze important data, ranging from the number of jobs to total core hours used against the walltime, job queue performance, hardware failure, etc. The dashboard also helps the HPC administrators to update the scheduler job queuing policy based on the historical data analysis of HPC jobs for better utilization of ANTYA resources.