Optimization of Service Migration Decisions in Mobile Edge Computing Based on Markov Decision Processes
Xiaoming Zhang, Zhan Quan Wen, Ran Tang, Xiaoke Wang, Hantao Liu, Miao He, Dehao Ren, Wenzao Li · 2024
In order to make optimal migration decisions promptly and accurately in the presence of uncertainty in service migration in mobile edge computing, we propose a service migration optimization model based on Markov Decision Processes (MDP) that abstracts the underlying spatial relationships between users and servers using the distance between them. Utilizing a value iteration algorithm based on differential equations, we further reduce the decision latency in service migration by treating the costs incurred in the decision-making process as value functions, calculating the minimum cost generated during service migration. Simulation results demonstrate that the decision time of this method is 0.1% of that of other traditional Markov decision models using standard value iteration and policy iteration. Moreover, compared with random migration and constant migration, the strategy obtained by this method is closer to the standard strategy obtained by using value iteration. This proves that the model has high efficiency and accuracy in the task of service migration in mobile edge computing.