| Cloud computing is an important part of China’s new infrastructure.Its application fields involve manufacturing,education,transportation,medical treatment and many other industries.With the development of 5G and other new generation information technology,the rapid growth of user demand has produced a large number of request workflows.How to quickly schedule workflows and reduce costs has become an urgent problem for cloud service providers.The container service cloud platform based on micro service and container technology has become its solution.Task scheduling and container scaling are the key technologies for building a container service cloud platform,which schedules tasks in the workflow to corresponding containers for execution,and scales containers according to requirements,thereby improving execution efficiency and reducing costs.At present,some researchers have carried out exploration and research on task scheduling and container scaling technology.However,the existing research work still has the following shortcomings: 1)Most of the existing workflow scheduling algorithms focus on static single workflow,do not consider the processing of real-time multi-workflow,ignore the dynamic characteristics of workflow and multi-workflow execution Uncertainty of timing,causing it to exceed workflow deadlines and increasing execution costs.2)Most of the existing container scheduling algorithms only consider a single factor such as CPU or memory,ignoring factors such as the pull time of the container image,the relationship between the container and the virtual machine node,and the execution association between different containers,resulting in low resource utilization,Poor quality of service.In response to the above problems,this thesis conducts in-depth research on the related theoretical technologies of task scheduling and container scaling in cloud platforms,and proposes a container-based real-time multi-workflow elastic scheduling method and a multi-objective optimization-based container scheduling method.In this case,the success rate of the scheduling algorithm is improved,and the cost of resource rental is reduced.The main work and contributions of this thesis are as follows:In view of the dynamic characteristics of workflow and the uncertainty of multi-workflow execution time,this thesis adds task pool and ready queue to the task scheduling of microservices,and uses them to coordinate ready tasks and subsequent waiting tasks to reduce the fluctuation of current task execution time.The impact on subsequent tasks can reduce the uncertainty of real-time multi-workflow? a method that combines task scheduling of microservices with container scaling is proposed.This method adopts a heuristic task scheduling algorithm based on urgency,considering The impact of container images on task execution,determine the scheduling scheme,and the type and number of required containers to provide support for container deployment.The simulation results show that the workflow scheduling algorithm ESRW proposed in this thesis reduces the resource rental cost by 25.9% on average compared with the similar algorithms ESMA and Pro Li S.Aiming at the problems of high cost and poor service quality caused by a single factor in existing container scheduling algorithms,this thesis considers six key factors that affect container performance: correlation between containers,correlation between containers and nodes,and nodes pulling containers Mirroring time,the number of containers in the node,CPU and memory utilization,a container scheduling method based on multi-objective optimization is proposed.This method establishes a scoring function for these factors,and combines the TOPSIS algorithm to find the most suitable node deployment for the container,which reduces the cost of resource rental and improves the service quality of the application.The simulation results show that,compared with the similar Spread algorithm and Random algorithm,the container scheduling algorithm MOCD proposed in this thesis reduces the scheduling time by 9.7% on average,and increases the maximum TPS of the container cluster by 12.5%.It effectively reduces the container deployment time and improves the execution efficiency of the microservices in the container.This thesis applies the improved algorithm proposed above to the container service cloud platform,designs and implements the Kubernetes scheduler component,integrates it into the Kubernetes cluster,and schedules and manages the services deployed in the cluster,which proves the feasibility of the proposed scheme and the practicability of the system. |
