Optimal Design And Implementation Of Migration Strategy For Computing Services In Mobile Edge Computing Environment

With the rapid development of information and communication technology in recent years,computing-intensive services such as AR(augmented reality),VR(virtual reality)and face recognition have developed rapidly.When these applications which need a lot of resources are implemented in mobile devices with limited computing and battery capacity,it will generate high repose and large energy consumption.In the mobile edge computing environment,considering the dynamic changes of network bandwidth and battery capacity of mobile devices,this paper designs a service migration strategy model based on deep reinforcement learning,which can make service migration decisions to complete computing tasks with less comprehensive cost according to the factors of network bandwidth and battery capacity of mobile devices,and applies the model to a system based on face recognition service.The specific work of this paper is as follows:In this paper,a service migration algorithm model based on deep reinforcement learning is designed and optimized.First of all,the related technologies based on mobile edge computing environment are studied,including mobile edge computing technology,service migration technology,Docker container technology,and deep reinforcement learning technology.Secondly,this paper studies the impact of two key dynamic factors on the service migration overhead in the mobile edge computing environment.Thirdly,this paper proposes a service migration algorithm model based on deep reinforcement learning.Finally,a service migration simulation system simulating mobile edge computing is built by using the python language simulation environment.By continuously optimizing the parameters in the deep reinforcement learning system,For example,the size and learning rate of experience playback take the comprehensive cost of service delay and energy consumption of a mobile terminal as the optimization objective;the results are compared with the greedy algorithm and dynamic planning algorithm,and the results show that the strategy obtained by the proposed service migration algorithm has better delay and energy consumption performance.A service migration system based on a face recognition application is designed and implemented.Using Tensorflow Lite open-source framework,the service migration algorithm model based on deep reinforcement learning,which is optimized and trained by parameters,is transformed into a file format that can work on mobile devices and implanted into face recognition application.On the server-side,the Docker container technology is used to design and realize the automatic configuration of face recognition service,automatic processing of configured face recognition service,and automatic return of calculation results.Under different network environments and different power levels of mobile devices,the response delay and energy consumption of mobile devices caused by face recognition service are measured,and the performance of the migration algorithm model proposed in this paper is verified.The experimental results show that the proposed service migration model based on deep reinforcement learning has better performance,which can save 39.13% time delay and 48.67% energy consumption compared with the service computing on the mobile device side,34.35% processing time delay and 52.65% energy consumption compared with the service computing on the server-side.