TASC (Tuning Applications for SuperComputers)
TASC is a software solution developed at RCC MSU aimed to help any HPC user in detecting performance issues in their jobs, finding root causes and determining possible ways for their elimination.
TASC consists of 3 subsystems. The first one automatically performs primary analysis of all jobs in order to detect possible performance issues and promptly notify users about them. The second subsystem is aimed at detailed analysis of a particular program, using the information about possible root causes of efficiency degradation obtained during primary analysis. The third one helps system administrators and management of supercomputer centers to study efficiency of supercomputer SW/HW components functioning.
The results of primary performance analysis are currently provided via Octoshell management system (“Efficiency” tab) for Lomonosov-2 job flow.
1. Shvets P., Voevodin V., Zhumatiy S. Hpc software for massive analysis of the parallel efficiency of applications // Communications in Computer and Information Science. — Vol. 1063. — Springer International Publishing, 2019. — P. 3–18. DOI: 10.1007/978-3-030-28163-2_1
2. Shvets P., Voevodin V., Zhumatiy S. Primary automatic analysis of the entire flow of supercomputer applications // Proceedings of the 4th Ural Workshop on Parallel, Distributed, and Cloud Computing for Young Scientists. — Vol. 2281 of CEUR Workshop Proceedings (CEUR-WS.org). — 2018. — P. 20–32.