I/O in Machine Learning Applications on HPC Systems: A 360-degree Survey

arXiv:2404.10386v1 Announce Type: cross
Abstract: High-Performance Computing (HPC) systems excel in managing distributed workloads, and the growing interest in Artificial Intelligence (AI) has resulted in a surge in demand for faster methods of Machine Learning (ML) model training and inference. In the past, research on HPC I/O focused on optimizing the underlying storage system for modeling and simulation applications and checkpointing the results, causing writes to be the dominant I/O operation. These applications typically access large portions of the data …

abstract applications artificial artificial intelligence arxiv computing cs.ai cs.dc cs.lg demand distributed excel faster hpc inference intelligence machine machine learning machine learning applications performance research survey systems training type workloads