Research On P2P Traffic Identification In Asymmetric Routing

P2P applications, such as network media-on-demand (MoD) and file resource sharing, havebecome an essential in our network lives because of high development of P2P technology. P2Ptraffic has taken up a great much bandwidth, which has brought many problems such asQoS(Quality of Service) of traditional traffic in no guarantee. To be worse, the exhibition formatof P2P traffic is changing from bidirectional flow to unidirectional flow gradually with thetendency of asymmetric routing boosting from edge network to core network in topology. A newtechnique is urgently required for ISPs(Internet Service Providers) to analyze, monitor, manage,and account P2P traffic in asymmetric routing. Therefore, research on P2P traffic identificationin asymmetric routing is of great realistic significance.Supported by “High Creditability Network Traffic Management and Control System”project of National High Technology Research a nd Development Program of China(863Program), a dualistic method and a fine grained method of identification of P2P unidirectionaltraffic are proposed separately from the asymmetric routing. A real-time P2P unidirectionaltraffic identification prototype system is designed and implemented by the combination of thesetwo methods finally.Main work and contributions of this dissertation are outlined as follows:1. A new dualistic method of identifying P2P unidirectional traffic based on C4.5algorithmis proposed. The method aims to identify unidirectional traffic, which selects10fast computedtraffic features paying attention to both identification accuracy and identification in real-timebased on C4.5algorithm and is usable to all transport layers carrying P2P traffic. Experimentresults in five different kinds of asymmetric routing present: this method has high overallaccuracy and short identification time, which provides a path to dualistically identify P2P trafficin real-time under asymmetric routing.2. A novel method of identifying P2P unidirectional traffic in fine grained way based onHierarchical Weighted Semi-supervised Affinity Propagation(Hi-WSAP) algorithm is proposed.Hi-WSAP algorithm improves its converge speed greatly by pre-computing similarity matrixbetween samples in subset after hierarchical division, which also depends on clustering labelingalgorithm to acquire semi-supervised model and could be identify unknown P2P applications.Experiment results in five different kinds of asymmetric routing show: this method has highF-Measure and low time complexity, which satisfies P2P traffic identification in fine grainedway in asymmetric routing.3. A real-time P2P unidirectional traffic identification prototype system is designed andimplemented which can be used in asymmetric routing. The key technique and module of thisprototype system are introduced in detail for the demand of real-time and accuracy. Test resultsshow that this prototype system:(1)reaches the accuracy86%-94%and consumes89seconds for110,500unidirectional flows when to identify P2P in duality;(2)reaches the accuracy80%-87%and consumes105seconds for94,600unidirectional flows when to identify P2P in fine grainedway, which achieves the needs of High Creditability Network Traffic Management and Control System.