Research And Design Of Cloud Resource Management Method Based On Kubernetes

In recent years,cloud computing container technology has a great development,this technology can be easily applied to the development and deployment,because of its good advantages,so the application in the enterprise is more and more.Docker,was one of the containers is cloud computing technology of the present stage Docker users oriented operation is limited,mainly the lack of service deployment,resource scheduling,monitoring,alarm and other related operations,Kubernetes as another kind of cloud computing technology,the container just makes up for the Docker in these areas,it has the characteristics of can provide convenient for the majority of users.However,the default resource scheduling policy of Kubernetes is too simple and the algorithm is single,which cannot meet the needs of different users in the actual production environment.Therefore,after studying the core technology of Kubernetes,this paper proposes a user-binding preemptive resource scheduling strategy,and proposes its own modification measures for relevant algorithms and policies.The main research content of this paper is as follows:Improved the user binding strategy in Kubernetes cluster,studied and added the user binding matching rules,added the Pod Node resource description file attribute,which can more effectively match the target Node Node,and reduced the time required for Pod scheduling with the increase of the work Node;When the cluster work nodes are short of resources,low-priority Pod transfers resources to high-priority Pod.This strategy can reasonably allocate the resources of the host machine.The division of Pod queue priority is increased.According to the actual usage of CPU and memory,a priority queue is designed and implemented,so as to avoid the waste of cluster working time caused by frequent re-invocation of resource recovery pods and improve the efficiency of cluster operation.The dynamic load balancing strategy was designed.According to the Pod Node resource description file,the nodes to be scheduled were scheduled to the most suitable Node Node,and the Pod in the Node with high load was scheduled to the Node with low load to maintain the load balancing of the whole cluster.