Research And Implementation Of Application-Oriented Traffic Identification Algorithm

With the expanding usage of Internet and increasing bandwidth of network, there are a large number of newly emerging network applications, which have more complicated communication modes and traffic patterns than traditional applications and make it difficult to analyze the Internet traffic. This thesis analyzes the network traffic at the application level and concentrates on the application-level traffic identification and characteristic. We focus on the following key question, i.e. how to identify the traffic from individual application?Many traditional technologies can't adapt to the development of the situation and the traditional technology of network traffic analyse, especially the traditional technology of application-level network traffic analysis, face great challenge. Under this situation, this thesis gives research on the application-level traffic identification, which has great significance. This research is much valuable to many research fields, such as network planning, network problem detecting, network usage accounting, even anomaly traffic detection, and network traffic prediction etc, and gives a better understanding to the complex network traffic. In the mean time, this reasearch can also enrich the theory of network traffic analysis, and can identify and analyze major part of network traffic. The traffic monitoring and analysis is fundamental to network management, so this research plays an important role in propelling the theory of network management. The network manager can use the result of traffic monitoring and analyzing to do many researchs, such as performance management, fault managment, security management, etc.First, this thesis gives full analysis to the sFlow technology, and proposes a framework of network traffic analysis system based on sFlow technology, and then implements the module of traffic analysis and performance management based on the sFlow technology in YH0412 network management system which can capture, decode, analyze the sFlow datagram and fulfill the flow-level traffic identification and the network performance management based on the sFlow datagram information.Second, after the thorough analysis of nowadays algorithms of traffic identification, this thesis proposes an application-level traffic identification algorithm based on the flow relative relations. The key idea is that through careful exploration of relations between flows generated from the same application, the flows is grouped based on the relations and the flows which have been grouped into the same group can be seemed as flows from the same application. This algorithm doesn't give examination on payload of packet, and so its efficiency and applicability is better.We have implemented the kernel module of this algorithm. We have also evaluated this algorithm and analyzed the performance, and the results show that this algorithm performs well; and has better scalability; and on some condition the time spending is also better.