A Review of Resource Management in Fog Computing: Machine Learning Perspective

Abstract

Fog computing becomes a promising technology to process user's requests near the proximity of users to reduce response time for latency-sensitive requests. Despite its advantages, the properties such as resource heterogeneity and limitations, and its dynamic and unpredictable nature greatly reduce the efficiency of fog computing. Therefore, predicting the dynamic behavior of the fog and managing resources accordingly is of utmost importance. In this work, we provide a review of machine learning-based predictive resource management approaches in a fog environment. Resource management is classified into six sub-areas: resource provisioning, application placement, scheduling, resource allocation, task offloading, and load balancing. Reviewed resource management approaches are analyzed based on the objective metrics, tools, datasets, and utilized techniques.