Research On Server Selection Strategy Based On Multi-armed Bandit In Mobile Edge Computing

The explosive growth of data is the main theme in the Internet of Everything(Io E)era.As the growth rate of network bandwidth is far from keeping up with the growth rate of data,it is accompanied by the emergence of a large number of new applications.High demands follow.Mobile Edge Computing(MEC)is a new computing model for computing at the edge of the network.Edge servers are deployed near mobile terminals to provide cloud computing capabilities to reduce the task delay of mobile terminals.Although the set of edge servers deployed close to the mobile terminal improves the quality of service for users,a key issue is how the mobile terminal selects the most suitable edge server in the edge server set to fully realize the maximum potential of mobile edge computing.At present,many scholars have done research on this,but there are still shortcomings:(1)The uncertainty of the environment and the lack of relevant prior information are not considered when selecting the edge server to uninstall,(2)In the scenario of multiple edge servers,Considering the different performance of the edge server,the computing resources on the edge server cannot be utilized to the greatest extent.In order to solve the above shortcomings,the paper mainly conducts research from the following two scenarios:(1)In the single-user multi-mobile edge server scenario,a dynamic edge server selection and uninstallation scheme is proposed.Assuming that the edge server's CPU computing resource changes meet a certain normal distribution,first establish the edge server task completion probability The model,offloaded to the edge server and offloaded to the remote cloud server will feedback the corresponding return value to the user,and solve it by establishing a multi-armed slot machine model to maximize the total return value of the user.Secondly,it proposes an offloading strategy for edge server selection based on multi-armed slot machines.The simulation results show that the proposed algorithm is better than the random selection and average resource allocation algorithms in terms of the total user return value and the total delay of user task execution.(2)In the multi-user and multi-mobile edge server scenario,the uncertainty of the task reaching the edge server is more specifically considered.First introduce the M/M/1queuing model in the queuing system to build the edge server queuing model.Secondly,combined with the exploration and utilization of the multi-armed bandit problem,according to the return value obtained by the user every time the edge server is selected and continuously updated,with the goal of maximizing the total user return value,a multi-armed slot machine-based edge server selection offloading strategy is proposed.The simulation results show that the proposed algorithm is better than the random selection and average resource allocation algorithms in terms of the total return value of all users and the total delay of all user tasks.