Research On Resource Scheduling Algorithm Based On Kubernetes Container Cloud

With the continuous development of the new generation of information technology,it has brought a great change to the Internet industry and prompted a new computing model.Traditional virtual machine-based virtual infrastructure has been widely used in cloud platforms,which has the problems of low resource utilization,low scheduling efficiency and inconsistent software stack environment.Operating system-level virtualization technology of Google began to emerge,in the cloud environment,containers represented by Docker began to be the cloud service providers as the underlying virtualization technology of cloud platform.The corresponding cluster management scheme Kubernetes has been widely used by the industry as a container cluster orchestration system.Then built a new generation of Paa S cloud platform,which has been widely used.This research first analyzes and studies the related concepts and technologies involved in the Docker container technology and Kubernetes container arrangement tools.On this basis,the Kubernetes resource scheduling module is deeply studied and analyzed,and the problems existing in the Kubernetes default scheduling strategy are pointed out.Then,in the application of different resource consumption types,a optimal scheduling strategy based on Pearson correlation coefficient method is proposed.The method selects the last host node for Docker container according to the similarity between the Docker container and the Node node,which can effectively improve the resource balance of the node.Then the automatic scaling mechanism of Kubernetes is analyzed and studied,and the problems existing in Kubernetes default automatic scaling strategy are analyzed,an optimization scheme is proposed.To solve the problem of response delay in the existing scaling mechanism,a load scaling algorithm based on load prediction is proposed.The grey prediction method and average autoregressive combination prediction model are used to predict the load applied in web cloud platform.Adjusting Pod copy according to the predicted load and response load dynamics,expanding capacity before peak load,and the load scaling of users decreases.Finally,the two optimization strategies are verified and analyzed,by building Kubernetes distributed cluster.Experimental result shows that the optimal scheduling strategy based on Pearson correlation coefficient method can effectively improve node resource balance and reduce resource fragmentation.The scaling strategy based on load prediction can effectively reduce the response time and improve the service quality of the cluster.