Packet-analysis based flow identification system for Peer-to-Peer application

Recently, Peer-to-Peer (P2P) streaming occupies a large number of network bandwidth. To keep network bandwidth from being occupied by a certain kind of network service, the traffic generated by the service should be limited. The limitation operation often closely depends on flow identification. This paper presents a P2P flow identification method on the basis of deep-packet analysis. The proposed method uses a distributed architecture, through client-side analysis of the statistical traffic characteristics, avoiding large amounts of analysis about packet payload on the network and parsing and restoring of the protocol. It also can be used to restrict the network bandwidth occupied by a large number of traffic on the server of the export gateway, thereby reducing the non-equilibrium bandwidth occupation. The verification experiments on the overall design show that P2P stream packets can be identified effectively by this system with more than 80% success rate.