System for Analysis of the Performance of Scientific Jobs in Distributed Systems

Modern high-energy physics experiments require a large number of computations for simulation, data analysis, and reconstruction. These computations are distributed across hundreds of thousands of tasks on various computing resources. To analyze the performance of these resources, an automated system has been developed by the authors, which collects and processes data about resource usage and CPU core performance. This data is visualized through a web application, showing the relationship between execution time and CPU core performance. The system is used for analyzing tasks from BM@N, MPD, SPD, and Baikal-GVD experiments within the DIRAC system at the Joint Institute for Nuclear Research.