P2P Scalable Video Streaming based on Imprecise Computation Scheduling

P2P network can be employed in a video streaming system to improve the system scalability with lower cost. To achieve jitter-less video playback on the requesting peer, it requires both the requesting policy on the requesting peer and the scheduling policy on each sending peer. Earlier works for dedicated streaming servers, however, are not suitable for P2P streaming. In this thesis, we propose a resource management mechanism for the sending peers in P2P scalable streaming. The workload model and real-time scheduling algorithm are based on imprecise computation. The resource allocation algorithm exploits the utilization of available bandwidth by considering data dependency and maximizing the utility enhancement of optional tasks. We use test clips encoded in H.264 SVC in the experiments. The results show that our algorithm utilizes the available bandwidth and the gain in playback quality is better than dual priority scheduling algorithm when the bandwidth is insufficient to receive the whole high quality stream.