Font Size: a A A

Research And Implementation On P2P Traffic Identification And Control

Posted on:2012-07-30Degree:MasterType:Thesis
Country:ChinaCandidate:D ZhaoFull Text:PDF
GTID:2218330338963036Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Recent years, as a new model of Internet applications, P2P technology has become a hottopic in the world rapidly with its services having occupied more than 70% of the total internettraffic. In pressure of these huge flows, it is a great challenge for ISPs to achieve more effectivemanagement, including the realization of effective P2P traffic identification and control.The thesis proposes an identification model based on the ensemble learning algorithmintegrateing DTNB,ONER, with the BP neural network algorithm into one .The model usesnetwork flow characteristics as well as the classification algorithms of machine learning.whilethe identification process includes three steps,namely ,to obtain network flow characteristics, toselect the feature of P2P traffic ,to establisht the flow classification model. The model combinesthe N-fold cross-validation with test sets to achieve the rationality and effectiveness of thepresented method. The traffic classification model has higher P2P flow identification accuracyfor the experiment results show that average accuracy reaches at 97.27%At the same time, the thesis proposes a P2P flow control mechanism, PTCM(P2P TrafficControl Mechanism), based on the model of FARIMA self-similar, which can intelligently adjustthe traffic control strategy according to the burst the network traffic. Also a non-P2P trafficpacket loss rate is used to assess the rationality and the effectiveness for the presentedmodel. The model prove to be a effectively way to control P2P traffic for the experiment resultsshow that the average rate of the non-P2P traffic packet loss is 5%.Finally, a distributed P2P traffic detection and management system based on Netfilterfirewall framework on Linux platform is realized which can be installed on net-bridge. Themodel includes an engine that implement P2P traffic identification and PTCM, which providesthe management strategy of bandwidth, and sends the management strategy to net-bridge by themessage. In this case, net-bridge can allocate the bandwidth properly.The research result of the thesis is helpful for bandwidth usage and the dynamic resourcebalance. It can provide powerful and effective guard for the network's security so as to ensure ahealthy and harmonious environment.
Keywords/Search Tags:P2P, traffic identification, traffic control, machine learning
PDF Full Text Request
Related items