Microservice Dynamic Migration based on Age of Service for Edge Computing

In the production process of a smart factory, the change of orders will cause adjustment or movement of devices, leading to prolonging the service response time. Therefore, microservice dynamic migration is necessary to optimize service freshness for industrial applications. We propose an evaluation index called Age of Service (AoS) based on the concept of Age of Information (AoI) to evaluate the service freshness. An average AoS minimization problem for microservice dynamic migration is formulated considering delay, AoS, and network changes. Since the problem is a dynamic process without the final migration strategies until the end of time, we transform the original problem into a one-slot delay minimization problem and solve it by Dueling Double Deep Q Network (D3QN). Simulation results show proposed algorithm can achieve lower AoS compared with no migration and random migration methods.

I. Introduction

In recent years, industrial internet of things (IIoT) has reached widespread development and is widely used in intelligent manufacturing. In the production process of smart factories, due to the limited computing resources of production devices, compute-intensive applications (such as fault detection and image recognition) can be transmitted to servers that have abundant resources for computation. However, computing-intensive applications will generate a large amount of production data. If all these data are uploaded to a cloud server for calculation it will cause transmission delay, which will greatly affect the efficiency of service response and even lead to production accidents.