| With the rapid development of the Internet,the number of Internet users has increased year by year,which has promoted the development of network applications.Different applications and users generate large-scale network traffic every day,which brings a series of network management problems such as network performance degradation,network status monitoring difficulties,and network security risks.Higher requirements are put forward for the classification,statistics,query,and analysis of network traffic.By classifying and counting network traffic,network administrators can rationally configure network resources and provide users with customized services according to network operating status,as well as implement targeted traffic engineering and scientific and efficient network management.This paper has researched the related technologies of rapid classification and statistics of network traffic.The main work of the paper is summarized as follows:(1)Designed a three-dimensional sketch data structure DMatrix that can retain network flow graph structure information and support reversible query,which is used to quickly and accurately summarize graph streams,and can quickly count application flow statistics.Based on three real-world datasets,we have conducted a lot of experiments and analysis.The experimental results show that DMatrix can achieve accurate and efficient statistics and query of streaming data under acceptable storage resource consumption.(2)A lightweight real-time network traffic analysis framework App Sketch is proposed,which innovatively combines lightweight network traffic classification technology with sketch technology.It provides more fine-grained application flow data statistics than existing solutions and supports multi-dimensional query analysis.Experimental results based on real-world datasets show that App Sketch can achieve rapid classification and statistics of network traffic while maintaining fairly good performance in terms of query accuracy. |
PDF Full Text Request