Playout Adaptation and Pull-Push Hybrid Scheduler in Live Peer-to-Peer Streaming

Abstract: Together with the wide spread of network devices with broadband access, streaming has become more and more popular as the favorite mechanism for broadcasting real-time events to a large number of users. The traditional approaches to live streaming include IP multicast and client-server based streaming. Of them, IP multicast is not globally deployed while client-server streaming implies expensive bandwidth provision at the servers. Therefore, recently a lot of researches have been focused on peer-to-peer streaming solutions due to theirs ease of deployment, cost-effectiveness and scalability. In this thesis, we first describe and implement a novel algorithm which addresses the peer churn issue in live peer-to-peer streaming system. This algorithm changes the playback delay in order to keep the average loss rate within predefined values. The correctness of the algorithm as well as our implementation is then confirmed through experiments on PlanetLab. The second contribution of this thesis is an implementation of the current state of the art peer-to-peer streaming approach, Coolstreaming+, for future evaluations and improvements. The implementation is then validated through the relation between loss rate and playback delay as well as between loss rate and number of sub-streams.