Improving HPC applications scheduling with predictions based on automatically collected historical data
Author:
Fenoy García, Carlos
Other authors:
Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors; Corbalán González, Julita
Abstract:
This work analyses the benefits of a system in an HPC environment, which being able to get real performance data from jobs, uses it for future scheduling in order to improve the performance of the applications with minimal user input.
Subject(s):
-HPC -scheduling -performance monitoring -historical data -slurm -perfminer