Design And Implement Of Container Cloud Platform Based On Kubernetes

The emergence of cloud computing solves the problem of on-demand allocation and time-sharing of computer infrastructure resources,which greatly reduces the cost and technical threshold for enterprises to obtain computer resources.Containers occupy a large share of the cloud computing resource vrtualization market because of their fast startup speed,small footprint,and low environmental dependence.Kubernetes has excellent orchestration capabilities for containers(automatic expansion,service discovery,resource scheduling,etc.).Cloud platform solutions that combine containers with Kuberentes have been widely used,reducing enterprise operations and maintenance investment.However,during use,it was found that the Kubernetes-based Docker container cloud platform has the following problems:abnormal conditions in the Kubernetes cluster cannot be effectively monitored and alarmed,a single point of failure in the Kubernetes component causes cluster service interruptions to be unavailable,and the Docker container instance cannot be scaled Resources cause resource waste,Kubernetes-based continuous integration and continuous deployment systems are not complete,and there is no management operation interface that is not friendly to non-operation and maintenance personnel.In view of the above problems,this topic mainly studies from the following aspects:First,build a Kubernetes-based container cloud platform and improve the Kubernetes-based container cloud platform design,including the high availability and rapid upgrade of Kubernetes components,the automatic shrinking of Docker container instance resources,and the collection and cross-over of Kubernetes-based logs.Perform in-depth research on cluster management,networking solutions,and storage solutions.Second,design anomaly detection algorithms based on key performance indicators of machine KPIs(Cpu,memory,network,storage,etc.),design feature extraction methods for time series data based on statistical characteristics,contrast characteristics,and sliding windows,and build deep learning models Anomaly detection,compared with other models,has improved significantly on F1 measurement indicators.At the same time,a system based on the anomaly detection model is designed and implemented to detect anomalies in the cluster.Third,design and implement a continuous integration and continuous deployment system based on Kubernetes to improve the efficiency of development,testing,deployment,operation and maintenance.You can manage deployed applications,view the current machine resource usage,and view the results of anomaly detection,making the deployment process more convenient and reducing the complexity of deploying application services.More friendly to non-operation and maintenance personnel.Finally,the overall test of the container cloud platform is carried out,the effectiveness of the functions of the container cloud platform is verified by deploying the actual application services,and the performance of the container cloud platform is tested by using JMeter test tools.At the same time,the functions of anomaly detection algorithm,continuous integration and continuous deployment are verified.