Optimization Of High Performance Data Traffic Analytical System And Data Analysis

Although Internet is an abstract idea,it is treated as a kind of commodity and always influences our life-styles and work efficiency.Recent years,the rapid development of network has been witnessed,especially Mobile Internet.From the point of Internet Service Provider,it leads the burst of the traffic in the network.Not only the challenge it brings us,but also the opportunity of development,which let us ring in the "Big Data" era.However,to create value from Big Data,the first step is the collection of data.With the increase of network traffic,the network traffic monitor nowadays must be capable of processing the traffic data generated from ten million users.How to optimize the performance combined with accuracy of the monitor seems to have highly positive value on traffic monitoring and big data analysis.In this thesis,the topic is about the optimization of high-performance data traffic analytical system and user behavior analysis.Firstly,the basic framework of the analytical software system is introduced,which provides follow-up studies with technological basis.For the optimization on the data traffic analytical system,we focus on two aspects:performance and matching accuracy.In the chapter of performance optimization,with the help of the report outputted by the profiler Intel VTune,we detect that the performance bottom necks are caused by unfriendly hardware access.In the research on the traffic matching algorithm,we point some limitations of the existing algorithm and propose a new algorithm which can cope with the pipelining connection.Afterwards,based on the records from the system,we conduct research on two aspects:1).the relationship between users' mobility patterns and online behavior;2).the cross-town events.Finally,we conclude the achievement of our research and prospect for future studies.