Research Of Monitoring And Scheduling Technology On Kubernetes

Nowadays,the container technology Docker provides a more lightweight and easy-to-use deployment solution for applications,and the Kubernetes(K8s)has become the de facto standard for large-scale containerized application deployment.However,the Kubernetes scheduler does not take the disk I/O load of nodes into account,which leads three problems:(1)The scheduler takes the sum of the CPU and memory requests of the Pod on the node into regard,but not senses the actual resource usage of the node,which affects the accuracy of the scheduling result.(2)Multiple I/O-intensive applications may be dispatched to the same node,which cause I/O bottlenecks.(3)Pods are less likely to be scheduled on node with idle I/O and insufficient CPU,resulting in the waste of the node's I/O resource.To address these problems,we did work as follows.In this paper,we first empower Kubernetes scheduler to detect node disk I/O and CPU usage by using Prometheus.We design and implement a custom scheduling system based on resource monitoring.The system includes two modules: resource monitoring and custom scheduling.The resource monitoring component is based on the open source monitoring tool Prometheus,and the custom scheduling component is based on Kubernetes Extender.The custom scheduling component obtains node monitoring data from the resource monitoring component,and loads and runs the custom scheduling algorithm.Then we propose a dynamic scheduling algorithm named by Balanced-Disk-IO-Priority(BDI)and a dynamic scheduling algorithm called Balanced-CPU-Disk-IO-Priority(BCDI).To solve the problem of cluster disk I/O imbalance,we propose a dynamic I/O balancing scheduling algorithm BDI and do experiments to prove that the balance level of cluster disk I/O is improved by the BDI algorithm.To solve the issue that Pod cannot be scheduled reasonably due to the single resource bottleneck,this paper proposes the BCDI algorithm,which is a dynamic scheduling algorithm based on node CPU and disk I/O balancing.Also we do some experiments to illustrate the effectiveness of BDI algorithm on improving the node CPU and disk I/O balance level.And experimentally prove that the BDI Algorithm improves the balance level of cluster disk I/O,and BCDI Algorithm improves the node disk I/O and CPU consumption balance level and the overall disk I/O utilization of the Kubernetes cluster.