Enhancing Dependability of Fog Computing Using Learning-Based Task Scheduling

Fog computing (FC) has emerged as a promising platform for processing delay-sensitive Internet of Things (IoT) tasks. Fog nodes (FNs) are prone to failure, which necessitates the introduction of a mechanism to increase FC dependability. Leveraging a failure-aware task scheduling approach is one way to make FC a dependable platform for IoT task execution. However, the dynamicity, heterogeneity, and execution uncertainty inherent in FC make designing such a system difficult. To this end, we propose a learning-based dynamic fault-tolerant scheduling approach that considers the three intrinsic characteristics of the IoT-Fog environment to improve fog service reliability for time-constrained IoT tasks. The proposed approach provides a hybrid fault-tolerance solution based on proactive, reactive and replication failure handling approaches. It also integrates a dynamic task runtime and energy-awareness in the decision-making to address the dynamicity, uncertainties, and energy consumption challenges in the FC environment. To validate the performance of the proposed fault-tolerant task scheduling approach, we carried out extensive experimental analysis and compared it with several baseline approaches. The proposed task scheduling approach improves fog service reliability, energy efficiency, and resource utilisation compared to baseline methods.