Abstract The increasing P2P network traffic on the Internet has leaded to the problem of network congestion. In the consequence of the diversification of the P2P traffic and protocol, research on the management of P2P traffic has had many problems needed to resolve. P2P traffic Prediction is kernel problem in the P2P traffic management. Based on the P2P traffic characters, this thesis present a P2P traffic model, gived a traffic prediction algorithm bases on wavelet-analysis, and proved the accuracy of the algorithm. Simulation has experiment figures that the algorithm a high prediction precision and superior real-time performance.