Multi-Armed Bandit-Based Dynamic Control Method to Access Edge Servers

With the rapid expansion of Internet of Things (IoT) technologies, communication traffic has increased dramatically, leading to growing demands on cloud computing resources. Mobile Edge Computing (MEC) has emerged as a promising solution by deploying multiple edge servers near IoT devices to offload lightweight processing tasks. However, load concentration on specific edge servers could cause response delays, making dynamic load balancing essential. This paper proposes a dynamic control method to access edge servers in MEC environments using the Multi-Armed Bandit (MAB) algorithm. To realistically represent edge server processing, the proposed method adopts an M/M/1 queueing model. In the proposed method, each client independently determines an edge server to which it offloads its current task based on the MAB algorithm, which determines rewards based on past task processing delays. The proposed method introduces several reward determination strategies to reduce task processing delays. Through simulation experiments, we show the effectiveness of the proposed method.