Intelligent Service Deployment Policy for Next-Generation Industrial Edge Networks

Abstract

Edge Computing has appeared as a promising technology for realizing industrial computation data at the edge of the network. The fundamental challenge in edge-enabled industrial networks is how to deploy the service requests while utilizing the available edge resources efficiently. In this paper, we aim to design an intelligent service deployment strategy for simultaneously handling both Industrial Internet of Things (IIoT) generated dynamic service requests and edge resources in the next-generation industrial networks. Initially, we present the objective function as the mixed-integer nonlinear programming problem for optimizing the energy-delay trade-off in the edge environment. To accomplish this objective, we model a heuristic-based task execution strategy and exploit the advantage of Deep Reinforcement Learning (DRL) to make accurate decisions in industrial networks. The proposed DRL-based strategy can learn well to control the industrial networks from its own experience and guarantees to handle as many service requests as possible using the set of available resource constraint edge servers. Experimental analysis reveals that the proposed strategy is robust to network changes and achieves better performance than existing algorithms in terms of energy consumption up to 13%, delay minimization by 23%, and other Quality of Service (QoS) parameters.