Highlights on Utilizing Machine Learning for High Performance Computing Systems

Due to the emergence of data-driven Artificial Intelligence (AI) techniques in the domains of machine learning and deep learning, there is a growing trend of applying AI-based methods to address issues in high performance computing (HPC). AI techniques have demonstrated their potential to enhance HPC systems in various aspects, including power prediction, job scheduling, anomaly detection from HPC logs, and resource management. This paper explores how machine learning techniques are utilized to improve HPC systems. Specifically, it provides a comprehensive assessment of approaches that synergistically converge AI and HPC. The paper delves into the specific applications of machine learning in HPC, such as resource utilization, anomaly detection, job scheduling, performance variation detection, and HPC simulations. Additionally, it discusses the challenges associated with log data and the methods for online anomaly detection. By examining these areas, the paper highlights the significant impact of AI on enhancing the efficiency and reliability of HPC systems. The findings demonstrate that machine learning models can significantly improve prediction accuracy, optimize resource allocation, and enhance system reliability, leading to more efficient HPC operations.