Research On Multi-objective Service Composition Scheduling Method In MEC Environment

As a key technology for next-generation mobile(5G)wireless networks,mobile edge computing(MEC)is a novel cloud computing paradigm,it sinks services and functions that were in the central cloud to the edge of the network,so as to meet the key requirements of low latency and high computing power for computation-intensive applications.Under the MEC environment of heterogeneous cloud distribution,it is very challenging and useful to build a service composition scheduler with reasonable performance and computation speed.Service composition scheduling problem in edge cloud environment has been studied extensively.However,there are still some limitations in the existing research results.For example,most research work on scheduling policies usually focuses on a single index in user’s Qo S,while ignoring the reality that multiple performance indexes conflict with each other,which will lead to the decline of service quality and affect user experience.Therefore,it is appropriate to consider the scheduling of application services as a multi-objective optimization problem(MOP)to meet the demand of users and service providers.Preference-based methods can efficiently generate compromise surfaces in the target subspace that the decision maker is interested in,as a kind of preference-inspired co-evolutionary algorithms,PICEA-g sets up multiple random preference sets to co-evolve with the candidate solutions in the search process,which makes the candidate solutions that are difficult to compare horizontally between multiple objectives further comparable,and guides the candidate solutions to the Pareto optimal frontier.Therefore,it is very suitable for solving multi-objective service composition scheduling problems.Nevertheless,PICEA-g still suffers from high time complexity and local convergence problems.To this end,this thesis proposes an improved multi-objective coevolution algorithm called ch-LPICEA-g,which aims to maximize the reliability of services deployed on edge infrastructure while minimizing the overall service completion time and cost for users.It’s an efficient heuristic algorithm,where the logistic and tent maps as two chaotic systems are applied in generating chaotic values to overcome the permute convergence in the initial population and the genetic operators.Also,clustering is applied to perform evolution operations in local areas so that the performance of original PICEA-g can be improved.We considered heterogeneous cloud services under different resource configurations,then conducted simulation experimental studies based on multiple scientific and randomly-generated service composition templates and location data sets in real world.The experimental results clearly show that the algorithm performance of the proposed method possess obvious advantages over the traditional methods in all cases.