Research On Container Cluster Management Of Shield Monitoring System Based On Docker

With the continuous increasing of the complexity of business systems and the number of functional modules,the Microservice Architecture emerges as the times require,the flexible features of loose coupling can solve the complexity of application effectively.Container virtualization technology represented by Docker has gradually become the best practice method for Microservice Architecture.Container technology solves the problem of differences in application development,testing,and deployment environments,so that developers can focus more on the application itself.As the scale of container clusters growing continuously,manual maintenance and deployment of operation and maintenance personnel can no longer cope with the challenges brought by large-scale clusters.Thus,more efficient container cluster management platforms are needed to complete tasks such as container management,resource scheduling,and load balancing.The container orchestration technology represented by Kubernetes has been widely used in actual business production environments.However,the granularity of containers is smaller,the relationship between containers is more complex than traditional virtual machines,also the efficiency of container cluster resource scheduling needs to be further improved.How to manage container clusters more efficiently becomes an urgent problem,since the existing default scheduling strategy of Kubernetes cannot fully meet the needs of actual business.In this thesis,we search container cluster management of shield monitoring system based on Docker.Firstly,related technologies are introduced briefly,such as Docker container technology,Kubernetes container orchestration technology and Microservice Architecture,and deeply understand these design philosophy and framework flow.Then,container cluster resource scheduling relies on efficient prediction of server load.Treat the server load as a continuous time series,we predict the server load based on the ARIMA-LSTM combined model,and the model fitting speed is accelerated by parallel computing.The load log of shield monitoring system is collected as data set.In comparison with existing methods,the load prediction model based on ARIMA-LSTM exhibits higher prediction accuracy and better fitting results.In the cluster environment,the speedup of parallel computing is higher.In response to the single problem of Kubernetes existing resource model,the load forecast model based on ARIMA-LSTM is applied to Kubernetes,integrating multiple factors such as CPU,memory,storage,and network bandwidth to improve the limitations of the resource model,also for finding the most suitable deployment node.At the same time,we optimize for the scenario of deploying multiple copies under the same resource controller of Kubernetes,which reducing redundant calculations and improving scheduling efficiency.Due to the high scheduling failure rate of Kubernetes when there is high load and insufficient resource,the optimal nodes are screened through multiple priority strategies to implement a preemptive scheduling strategy to prioritize the stability of high-priority services.Finally,the shield monitoring system was restructured based on the Microservice Architecture,and the unified authentication and authorization of each sub-business system in the microservice environment was achieved through the Service Account.Each service in the shield monitoring system has its own persistent data,and there is strong isolation between microservices.Thus,the external event service model is used to separate the event service from the main business service to maintain data consistency in the system and reduce dependence between services.The shield monitoring system based on Docker implemented in this thesis currently runs well in the production environment,it enables efficient management of container clusters,and better meet the business needs of smart management and control of enterprises.Even under high load conditions,the cluster can provide a stable business environment and provide reliable support for the subsequent service development,delivery,deployment and iteration.