Time-continuous computing offloading algorithm with user fairness guarantee

Computing offloading is a potential avenue to reduce transmission delay by moving computing tasks from cloud to edge nodes. Due to the limited computing capacity of edge nodes, performing effective computing offloading strategy is challenging. To simplify the problem, it is often assumed that the timeline is discrete and offloading decisions are only made at the end of each time slot. However, the assumption results in decision waiting time and increases service delay. In this paper, we first formulate a time-continuous computing offloading problem without the assumption of discrete timeline. In order to reduce service delay while guaranteeing user fairness, the optimization objective is constructed as the alpha-fair utility function of average task delay rather than just average task delay. Then, we propose a novel algorithm by adjusting the reward in standard reinforcement learning to solve this problem, and also prove the convergence of our algorithm theoretically. To the best of our knowledge, this paper is the first attempt to study the time-continuous computing offloading problem with fairness. Evaluations show that the proposed algorithm has better performance in terms of service delay and user fairness compared to five baselines.