Research On Service-based Application Deployment In Edge Environment

With the development of technology,mobile applications can provide more diversified and powerful functions to meet the growing needs of users.However,users have more stringent requirements for network delay of such applications and pursue real-time response and feedback.With the development of edge computing,to ensure the quality of user service,application providers prefer to deploy services at network nodes closer to users to reduce network delay.However,the edge environment is very complex,and how to reasonably deploy services is a difficult problem.The edge servers that provide resources have heterogeneity and limitations,the users who use mobile applications have mobility,the services that need to be deployed have reusability,and there are mandatory relationships between services.In addition,from the perspective of application providers,deployment cost is a significate factor.Therefore,in the research of service deployment in edge computing environment,aiming at the three challenges of service redundancy deployment,user mobility and application with complex structure,there is an urgent need for a deployment scheme that can not only meet the cost constraints,but also ensure the quality of user service.Aiming at these three challenges,this paper has carried out the following research:(1)Budget-constrained optimal deployment of redundant services in edge computing environment.Firstly,this paper establishes a redundant service deployment model in heterogeneous edge environment,then defines it as a multi-objective optimization problem under given budget constraints.Secondly,the concept of service priority is defined,and the calculation method is given to determine the redundancy of services.The k-medoids clustering algorithm based on request frequency filtering is used to preliminarily select edge servers.Then an improved genetic algorithm based on priority is proposed to obtain a scheme that meets the cost constraints and better performance.Finally,experiments on real data sets show that the proposed method is superior to other existing methods.(2)Service deployment based on user mobility and service composition.Based on the previous work,this paper further considers the uncertainty brought by user mobility to the problem of service deployment and proposes a service redundancy deployment scheme generation algorithm,which can provide lower network delay with higher one-time request success rate.Firstly,this paper uses LSTM model to predict user's future trajectory,then obtains user's future request vector combined with the request information and service composition,and finally uses the double clustering algorithm to obtain a reasonable deployment scheme.Finally,experiments on real user data sets verify the effectiveness of the proposed method.(3)Optimal deployment of service-based application with complex structure.Based on the research of linear structure application,the application deployment problem with complex structure is further studied.Fully considering the relationship between the services,this paper first proposes a deployment algorithm based on service correlation.Then,combined with graph theory clustering,the complex relationship caused by service reuse is further studied,and a service clustering based on graph theory deployment algorithm is proposed.Finally,by establishing the service communication weight network,a weight based heuristic algorithm is proposed,which can continuously optimize the deployment scheme through iterations.Finally,in several groups of experiments,it is proved that the algorithms proposed in this paper have good optimization effect and higher efficiency.