In a video-on-demand streaming system, each user watch dufferent video frame according to their arrival time. In the thesis, the concept of multi-source and the forward error correction scheme are implemented to decrease the workload bandwidth of server and reduce the packet loss probability of each peer in the peer-to-peer video-on-demand system. Here, in order to share data to some peers arrive later, each peer must cache part of video data that recently viewed. Each new arrival peer needs to contact multiple sources who have initial part of video data to get video streaming data, and each source transmits different part of video streaming. Once receiving all of these partial stream, the peer will get completely video data. Simulation shows that a suitable preserve time of peers in the system can be used. Thus, the workload bandwidth of server used and memory used by peers can be reduced, too. Furthermore, the packet loss probability is decreased by the sources diversity and the FEC recovery.