| Informatization in universities and colleges has been continuously improved for more than two decades,so that campus network is being used in all aspects of teaching,research,and daily life in universities and colleges.Now that mobile devices rapidly emerge,teachers and students have even more equipments connecting to the campus network.As a result,the network usage is beyond traditional wired networking,but is more concerned device mobility.Along with the numbers of both users and their connected devices,users add their commonly wired locations: from dormitories,laboratories,offices and other traditional fixed locations to teaching building,libraries,student centers,gyms and other public areas.Thousands of wired and wireless user terminals,frequent network roaming,real name network authentications,abnormal network usages,etc,feature the university users' authentication data with Volume,Velocity,Variety,and Value,the typical four V's of Big Data.Therefore,data analysis and data mining upon those data are very importa nt.In this paper,we propose a storage and analysis framework based on Elasticsearch's distributed full text search,managing and analysing massive user authentication data from a university to recognize users' behaviour patterns on using internet.The framework was composed of 3 modules including big data construction,data analysis and data demonstration.The module of big data construction has the function of the raw data extraction,filtration and conversion.So all kinds of authentication data could be indexed separately based on Elasticsearch cluster,which finally generate the massive date of Campus Network access authentication.The module of data analysis has the function of data search and analysis.The analysis rule could be user-defined.The modle of data demonstration present results in tables and charts.As a application system of this study,a behavior analysis system is used to find out the network status,abnormal clients and other special events.Convenient query analysis makes the user behaviour analysis more efficient.Customized analysis results help network administrators to identify the network problems earlier than the normal users,so that they can schedule the network maintenance in advance or assist the user proactively.These data have been used to optimise the quality of the network,improve the user experience,help making network management decisions and assist student management.It has been proved to conduct good results.This experiments show that the system provides solution of long term data storage for massive data,and efficient data analysis and query upon it.Behaviour analysis realises the values from the data.The whole solution is feasible and efficient. |
