Design And Implementation Of Cluster Management And Monitor System Based On SpringMVC

The accumulation of data in the information age has stimulated the development of big data processing technology,therefore,the construction and management of cluster attracts increasing attention.At present,most of the cluster management and monitoring platforms are commercial software of C/S architecture and require the support of many third-party plugins.They have high cost and time-consuming install process.Also most of them do not have the function of cluster deployment.The thesis focuses on the intelligent deployment and management of cluster and designs a B/S-based system of cluster management and monitoring.The main research contents are as follows:(1)Cluster deployment steps are researched and service installation process is abstracted as well as the design of underlying service software library.On the basis of multi-thread,the configurable flow of the cluster automatic deployment program is achieved through SSH2 connection and remote shell.(2)Management process of cluster and service level are designed and implemented.Besides,a parallel service copy growth algorithm is proposed to accelerate cluster node expansion and the intelligent expansion of cluster service is realized through optimizing the assignment of node role based on the distribution weight vector updating.(3)Based on event trigger monitoring,this thesis proposes a cluster monitoring scheme by combine scheduled service polling with streaming data processing,and realizes the multi-dimensional monitoring of service,node and job with Hadoop native API.(4)The threshold-based monitoring alarm is designed,and the fault tree analysis method of depth-first search is used to realize the rapid fault location after the cluster monitoring alarm.(5)User,role and privilege are combined to design the user rights control system which achieves the operation control of button level with Apache Shiro.Experiment and analyze are performed to the function of cluster deployment and monitoring functions.Results show that the system performs well in small and medium-sized cluster management.