Design And Implementation Of Service Migration Scheme For Low-orbit Network Based On Edge Computing

In recent years,mobile communications have developed rapidly.With the official commercialization of 5G,the development of terrestrial networks has come to the bottleneck,and people can't help but set the goal on the integrated satellite-terrestrial networks.Among them,the low-orbit satellite(LEO)networks can solve the problem of insufficient coverage of the terrestrial networks and is a supplement to the terrestrial network.With the development of all countries in the world,the LEO networks in the world has begun to take shape and is still developing at a high speed.At the same time,there are some problems in the LEO networks,mainly including the problem of the dynamic change of the network topology,and the resulting mobility management problem has become a key and difficult point of the LEO networks.Whether the problem of mobility management can be solved well determines the future development prospects and application prospects of LEO networks.From the perspective of multi-access edge computing(MEC),this paper conducts a series of studies on MEC mobility management in the context of LEO networks.First,the development status and research directions of LEO networks at home and abroad were investigated,and the development prospects of MEC in this scenario were determined.Then,we investigated the current LEO mobility management and MEC mobility management solutions,and after evaluating the characteristics and shortcomings of the existing solutions,a LEO network architecture integrating MEC was proposed.On this basis,by combining the network model to analyze the communication process in this scenario,the problem of a large number of routing updates caused by network topology changes and the resulting problem of poor scalability of the LEO network are improved.Based on this network architecture,a MEC simulation platform with containers as the core was built,and the MEC service discovery,service offload,service migration and other functions were simulated,and the effects of cold and hot migration and cloud and edge computing were compared.Aiming at the problem of MEC mobility management,a solution to discretize the service process into a multi-migration decision-making process is proposed to reduce the processing difficulty and cost.For the computing service process of a single user,the predictable prior information of the satellite constellation is used,and the global optimal Viterbi method is adopted to solve the migration decision.Aiming at the resulting local congestion problem in the network,a solution using reinforcement learning is proposed.Finally,the performance of the proposed scheme is verified by simulation experiments.