A fault-tolerant workflow scheduling method on deep reinforcement learning-based in edge environment

Edge computing emerges recently and shows promising solutions to support and boost applications for the Internet of Things (IoT). Nevertheless, by considering a distributed edge server system consisting of multi-users with time-variability and mobility, which may cause applications to be stopped due to resource faults or link failures. We present here a method that directly aims to make the execution of workflows more reliable and efficient. The proposed and designed a fault-tolerant method first analyzed the allocation of a dependency-based task, then tolerates task failure on edge nodes using the primary-backup (PB) strategy, and finally applies state-of-the-art Deep Q Network (DQN) algorithm to obtain the optimise task scheduling scheme. To the end, we conduct a simulative case study on real datasets and compare ours with other fault-tolerant methods. The result shows that our approach improves the reliability of workflow execution in edge environment.