A machine learning approach to online fault classification in HPC systems

Open Webpage

Alessio Netti, Zeynep Kiziltan, Özalp Babaoglu, Alina Sîrbu, Andrea Bartolini, Andrea Borghesi

Published: 2020, Last Modified: 27 Jul 2025Future Gener. Comput. Syst. 2020EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Development of FINJ, a novel open-source fault injection tool for HPC systems.•Release of Antarex, a freely-available dataset of faults in an HPC system.•A machine learning model for fault detection designed for online monitoring data.•Faults can be detected online with very high accuracy and low overhead.