| In recent years,with unprecedented challenge of the increasing service and traffic demands from geographically distributed users,application service providers(ASPs)such as Facebook,Google and Netflix are being confronted with the flexibility and agility of their workflow applications in terms of scheduling and deployment.By adopting microservice architecture,the complexity of the system is greatly reduced and flexibility and extensibility are increased,it is easier to scale,remove and deploy system components,and increase the flexibility of using different frameworks and tools.ASPs use microservice architectures to divide high coupling and complex workflow applications into a large number of fine-grained,low-coupling,and lightweight microservice tasks.With the development of virtualization technology under cloud computing,Container as a service(CasS)launched by cloud service providers can easily deploy and run large-scale microservice workflows.This thesis studies the optimization method of large-scale microservice workflow scheduling based on CaaS,which has important theoretical significance and application scenarios.Based on the large-scale microservice workflow scheduling problem under the container cloud platform,taking into account the constraints of different microservice workflow deadlines,the goal is to minimize the total renting cost of computing resources.A mathematical model of microservice workflow scheduling based on CaaS is established.Based on this model,an algorithm of scheduling microservice-based workflows in CaaS(SMWC)is proposed.The algorithm consists of six phases: microservice workflows sorting,task slack time computation,task sub-deadline distribution,task scheduling sequence generation,heuristic Docker container placement and schedule solution improvement.In order to verify the effectiveness and efficiency of the proposed algorithm,The variance analysis technique was used to correct the relevant parameters of the algorithm to obtain the best combination of parameters.The proposed SMWC algorithm is compared with the benchmark algorithm FLTM,which uses Amazon ECS Fargate Launch Mode as a computing resource model,under different microservices workflows size and deadline constraints.Experimental results show that the SMWC algorithm proposed in this thesis optimizes the benchmark algorithm FLTM under the constraint of different scale or deadline. |
