| With the explosive growth of the Internet and the increasing power of the computers, new demands and technology boost P2P streaming systems up in recently years. This application has been becoming one of the most promising services in emerging broadband integrated service digital network.Typical P2P streaming system consists of 2 subsystems, including live subsystem and VOD (Video-On-Demand) subsystem, but the early version supports live subsystem only until 2006. The VOD application gives more attractive interface and controllable watching time.Many researchers have focused on P2P Live streaming system, while, P2P VOD system is quite a new topic and lack of deep exploration. Most of exiting work on VOD system is concentrated on the protocol design of topology and the analysis of simulation results. Measurement based on VOD system appears until last year.In this paper, after a brief description of VOD system and outline of our reverse analyzing to the protocol and message format of PPLive VOD system, users' playing roles in system and watching behavior are studied through users' Buffer Map packets. Based on this method, we find out that the Contributors (upload only users) are much more than the Watchers, and Smoothers (smoothing watching users) are also more than the Jumpers (random watching users) in VOD system. It is quite different from currently P2P file dissemination systems (such as BT and eMule). The relationship between users' watching behavior and network sharing is revealed in this paper. It is obvious that the Jumpers provide more chunks and more rare chunks for sharing than Smoothers do, and Jumpers can do more contribution to the network sharing than Smoothers do.With the understanding of the current VOD system, our research in this paper gives a suggestion that better user experience can be expected by improving the protocol and designing in VOD system, according to the finding that network sharing is tightly related to the user watching behavior. A good VOD system should be applied to any kind of watching behavior habits including random seeking and smoothing watching. |