| With the rapid rise of network bandwidth and the all-round development of network applications, traditional offline or online network traffic detection and analysis system could not meet the current network monitoring requirements of high-speed network data processing, diversified network anomaly detection, historical and sensitive information quotation, time-correlated data detection, and other network applications. Based on data stream management technologies, the problems of data processing in network data monitoring system are discussed in this thesis.Researches and works included in this thesis are listed as follows:First, new features of data stream management technologies are analyzed, the typical DSMS prototypes and the key technologies in their implementation are studied and compared, which establishes the foundation of researches and innovations for the thesis.Second, an Entire Time-Data Stream Management System (ET-DSMS) architecture is designed for the problems of detection of time-correlated data and quotation of historical information in network monitoring applications. Three executions of query plans in ET-DSMS architecture are presented by expansion time-based cycle sliding window based on CQL continuous query language.Third, optimization technologies on system performance are studied, Feedback Control and Semantic based Adaptive load shedding mechanism, and Adaptive conditional-Filters Ordering Schedule strategy are proposed, and optimization strategies on ET-DSMS architecture based on the ideas of load shedding and adaptive schedule are given.Finally, the ET-DSMS architecture and optimizing mechanisms described above are applied to network data monitoring system. By researching the system operation process from functional structure, hardware environment and implementation process, the validities of designs of ET-DSMS architecture, load shedding mechanism and adaptive schedule strategy are proved by experimentations. |
PDF Full Text Request