Research On Time Sensitive Service Placement Strategy In Mobile Edge Computing

Mobile edge computing is one of the core technologies of the 5th generation mobile system.The large bandwidth,low delay and low power demand of 5G network equipment bring great challenges to the existing mobile cloud computing architecture.Mobile edge computing is the extension and supplement of mobile cloud computing,by processing some services on the network access side,data transmission through core network becomes unnecessary,which relieves problem of quality of service reduction and the transmission bandwidth pressure of core network,improves service continuity and service quality.Firstly,this thesis introduces the main characteristics of 5G communication and mobile edge computing.Through analyzing the shortcomings of mobile cloud computing,the necessity of mobile edge computing is introduced,after that,research background,application scenarios and current research hotspots are introduced.Furthermore,the service placement in mobile edge computing is described,and the existing research works are analyzed.Secondly,this thesis proposes an edge service placement strategy based on distributed deep learning.Firstly,in order to minimize the delay of all users' service requests and the cost of weighted service placement,the problem is modeled as mixed integer nonlinear program.Secondly,suppose that edge service placement strategy is determined,the optimal computing resource allocation scheme of edge-cloud is solved and proved by using convex optimization theory.Finally,the distributed deep learning is used to solve the problem of edge-cloud service placement.The theoretical proof and simulation results show that the proposed strategy can effectively reduce the delay of users' service request and the service placement cost of application service providers,and gradually approach the global optimal service placement strategy.Furthermore,this thesis proposes an edge service placement strategy based on improved genetic algorithm.Firstly,the service delay,resource utilization and energy consumption are modeled.Then,the multi-objective optimization problem is proposed.Finally,the decision candidate sets are obtained by using the improved genetic algorithm,the best solution of the candidate sets are selected by using multiple criteria decision making and technique for order preference by similarity to an ideal solution.Simulation results show that,the proposed strategy can effectively reduce the system delay,improve the utilization of resources,and save energy consumption.Finally,the research work of this thesis is summarized,further research objectives are proposed and the future research hotspots are prospected.