Design And Implementation Of Operation And Maintenance Monitoring System Based On Time Series Analysis

With the rapid development of the Internet,the number and variety of network resources and application services have increased dramatically.In order to centrally manage network resources,understand the operating status of network application services,and uniformly standardize the task scheduling process,an operation and maintenance monitoring system came into being.In the operation and maintenance monitoring system,various types of KPI data are mainly time series.At present,most operation and maintenance software lacks in-depth research on time series processing and analysis methods.Therefore,based on the development of current operation and maintenance technology,this paper selects servers,application services,and My SQL databases as the monitoring objects based on actual business requirements,and designs a type of operation and maintenance monitoring system based on the B / S architecture.And real open source data sets are used as experimental objects to study seasonal analysis of time series,trend prediction and real-time anomaly detection.The main research work of this dissertation is: 1.Research on the development model,overall architecture and database design of this system;2.In order to deal with the diversification of actual business scenarios,each type of monitoring object has implemented two data collection methods;3.Based on the exponential smoothing model,the missing points are complemented and the noise reduction processing function is implemented.The model is simple to implement and retains key features such as seasonality of metadata;4.A time series seasonal analysis mechanism based on a combination of autocorrelation analysis,fast Fourier transform,and STL decomposition is proposed.This mechanism can calculate the length of multiple seasonal cycles included in the time series;5.Aiming at the shortcomings of the HoltWinters model in the prediction of seasonal time series,a decomposition prediction model combining the MSTL decomposition algorithm and the exponential smoothing model is proposed,and the validity of the model is verified based on the real dataset;6.Aiming at seasonal time series,a class of real-time anomaly detection mechanism based on prediction model is studied.Based on the above analysis and research,compared with the traditional operation and maintenance monitoring system,the performance of this system in data processing and analysis has been improved.At the same time,this article looks forward to the future development of the operation and maintenance monitoring system.