Abstract: Peer-to-Peer(P2P) live streaming systems have recently shown great potential attractive on the Internet. However,large-scale deployment of such systems relies too much on their efficiency in dealing with high dynamic chan- ges, especially in the flash crowd period. hhe main reason is that the expansion of a P2P live streaming system largely depends on the of the time demand of streaming media applications. Based on the proposed analysis and experiment, we found the inherent relation between the system scale and the time, as well as the constraints. We constructed a generic model for P2P streaming system to concentrate on analyzing the process that the nodes added to the system in the flash crowd. The paper first argued that simply using the "supply-demand" concept model to describe the system scale is not enough, and then introduced the upper bound of system scale over time with a random partner selection strategy like Gossip protocol. Finally we showed the impact on system scalability by critical factors clearly through a comparative a- nalysis on Matlab R2010a platform.

Key words: P2P streaming system, Flash crowd, Scalability, Random partner selection strategy