Improving HPC applications scheduling with predictions based on automatically collected historical data

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