Design And Realization On DPI-based Network Traffic Monitor System

With the network infiltrating into the life more and more comprehensive,the network has become an essential part of life,so bandwidth and network service quality has become the focus of the telecommunications industry.With the network globalization,Skype,MSN and others harvest a large number of customers in a short time when access into China,IP talk has a rapid growth.Domestic and other companies have also seen a good prospect of development,and have followed,such as the Tencent Inc micro channel team using QQ in China's vast market has changed the way people's social communication.VoIP call erode the voice of the telecommunications business,formating a large shunt for the traditional telephone.All these problems have brought new challenges to traditional telecom operators.At the same time,when people use the network to form a huge data,analysis of these data can get valuable information.In this paper,we design an effective Packet Inspection Deep(DPI)monitoring system based on the research and analysis of traffic identification and control technology.In this paper,a method of combining DFI with traditional DPI traffic identification system is designed.Mainly for the current network traffic based on the HTTP video traffic,based on VOIP video calls applications and P2P based traffic identification and control,the demand for coarse grain reached a very good results.In addition to the system can also control some IM chat applications such as QQ,information transmission such as network disk and other software to,and have good effect with identifying and controlling.In this paper,the design of the identification and control scheme is presented in detail.Finally,the relevant functional and performance test results are given.The work done in this paper mainly includes the following aspects:(1)An identification model is designed.Studied several kinds of present popular traffic identification technology,a summary of the advantages and disadvantages of comparison,design a kind of containing the DPI and DFI two recognition system.(2)The research pattern matching algorithm,in order to improve the system efficiency of pattern matching.Carried on the thorough research in terms of matching algorithm,the classic AC multimode matching algorithm introduces the matching principle,the principle of decision tree C4.5 algorithm of machine learning fully understand,especially classifier classification algorithms and feature selection are detailed in this paper.(3)Research different types of network traffic characteristics.Combination Line,YouKu and so on case study the principle of protocol and network structure,the different matching mode.(4)Study Deployment plan of the system deployment,compared the advantages and disadvantages of different deployment structure.We design a flow recognition based on Linux Netfilter technology and overall architecture of the control system,according to the custom procedures to deal with traffic,has reached the control demand.USES the straight road to deploy,we design a series in one straight road network deployment structure.