Design And Implementation Of Cluster Monitoring Predicting Platform Based On Zabbix

With the rapid development of Internet and computer technology,cluster technology has been widely used for explosively growing network data.With the successful construction of clusters and the continuous expansion of nodes,the reliability and stability of the cluster itself has become an important factor that can not be ignored,so all kinds of monitoring systems have emerged.For SMEs,it is difficult to withstand the high cost of commercial monitoring software,and there is little ability to independently develop monitoring software.Therefore,it is necessary to provide a complete cluster monitoring service mechanism for SMEs.At the same time,in the existing monitoring systems,most of them adopt the “post-warning” approach,from the perspective of the management personnel of the cluster,it is possible to provide early warning before the system failure,which is beneficial to avoid major accidents.Therefore,adding a short-term real-time forecasting algorithm to the monitoring system and making it possible to perform an early warning based on the prediction result is particularly important.This article first fully understands the existing monitoring systems.After comparison and analysis,it realizes a real-time monitoring and forecasting platform for small and medium-sized enterprises based on the open source monitoring software Zabbix.The monitoring and forecasting platform monitors all kinds of indicators in the cluster in real time and predicts important performance indicators in real time.When monitoring or predicting abnormal conditions,the monitoring personnel are alerted or alerted,which is helpful for managers to know the operation status of the cluster in real time.The main work of this paper includes the following aspects: Based on Zabbix,a cluster monitoring solution is designed for small and medium-sized enterprises to collect cluster performance data in real time and master operational conditions;Study and analyze the data characteristics of cluster performance indicators,design a prediction model based on time series forecasting methods,and design and implement a short-term real-time prediction algorithm based on the prediction model;Analyze and modify Zabbix's open source code,add forecasting algorithm to the monitoring forecasting platform,and realize the short-term forecasting function of monitoring indicators;Based on the monitoring of the cluster,the automatic alarm and early warning of the fault can be realized through the script;Improve database performance by partitioning your database tables and creating automatically maintained partitioned stored procedures.After implementing the cluster monitoring and forecasting platform,this paper has performed a full functional test on the system and found that the system has sufficient accuracy,can fully monitor the cluster performance metrics,And when an abnormality occurs or is predicted,it can effectively perform warning or warning.At the same time,after realizing the short-term real-time prediction algorithm,By comparing and analyzing the actual and predicted values of the monitoring data,the validity of the prediction algorithm is verified and the purpose of “pre-warning” can be achieved.The results prove that the cluster monitoring and forecasting platform can ensure stable and robust operation of the cluster.This article completed the expected work.