Design And Implementation Of Metric Management And Analysis System For Unified Operation And Maintenance

Currently,technologies such as cloud computing,big data,and artificial intelligence are developing rapidly,and the national information construction is gradually deepening.Information systems have become an important part of the core competitiveness of enterprises.As a guarantee for the stable,secure,and efficient operation of information systems,IT operation and maintenance have received increasing attention.In the process of enterprise IT construction,multiple equipment and service providers are often procured to build various information systems.However,the network management platform or monitoring and operation system provided by equipment providers can only manage their own devices,resulting in problems such as multiple and miscellaneous operation tools,inability to unify management and analyze operation data,and high operation costs with low efficiency.Aiming at the above problems,this thesis first researched and analyzed the current situation and mainstream metric forecast and warning algorithms of well-known network management platforms at home and abroad,and then carried out requirement analysis,architecture design,function design,and system implementation for the metric management and analysis system.The system is divided into data access module,data storage module,metric management module,metric analysis module,and data visualization module,with the data access module and metric analysis module as the core parts of the system.To support the unified access of operation and maintenance data from multiple network management platforms,we proposed a modular ETL process construction method.Users can construct ETL processes through component drag and drop,and a field mapping method based on core words is proposed to improve the automatic unified mapping ability of data fields from multiple platforms.To solve the problem of forecasting and warning multiple types of metrics,we proposed an metric warning algorithm based on residual dynamic baseline.We constructed a multi-model integration predictor to achieve metric trend forecast,generate a forecast baseline,and then use N-sigma algorithm combined with Gaussian white noise to construct a residual dynamic baseline based on the forecast baseline to achieve metric warning.Experimental results show that the proposed metric warning algorithm is feasible and efficient.Compared with traditional 3-sigma algorithm and fixed baseline algorithm,it has higher flexibility and universality,and the warning effect is also superior.Finally,functional testing and performance testing were conducted for the metric management and analysis system to verify that each functional module achieves the expected effect,and the system can operate stably and efficiently,ensuring the provision of good metric management and analysis services for operation and maintenance personnel.