Machine Learning for Predicting Infrastructure Faults and Job Failures in Clouds: A Survey

Fault prediction in cloud environments is a critical task for ensuring the reliability and availability of cloud services. Machine learning (ML) techniques are increasingly used for this purpose due to their ability to recognize and predict patterns that may indicate potential faults. In this survey, we propose a taxonomy for ML-based fault prediction work, providing a critical overview, and evaluating the work from an algorithmic perspective. This includes identifying five key requirements for fault prediction in clouds and using them to evaluate the work. In this evaluation, we gain insight into the literature's current state and identify research directions. Our evaluations indicate that adaptability and interpretability requirements are not adequately met in the current literature, and future research could focus on developing fault prediction solutions that are adaptable to dynamic cloud environments, and can provide interpreta-ble explanations for their predictions.