P2P Live Streaming: Practice and Theory

Опубликовано 7 сентября 2016, 16:30

With the widespread adoption of broadband residential access, live video streaming may be the next disruptive IP communication technology. As an indication of the potential of live video streaming, recently a commercial P2P streaming system broadcasted the 2006 Chinese New YearΓÇÖs celebration to over 200,000 users, generating an aggregate bit rate in the vicinity of 100 gigabits/sec. CoolStreaming, PPlive, and ppstream are among the most successful deployments, all of which use pull-driven meshed-based designs. In the first part of this talk we will describe the general architectural characteristics of pull-driven architectures. We will then summarize the results of a recent measurement study of PPLive. As part of the measurement study, we have developed a PPLive crawler, which is used to study the global characteristics of a large-scale P2P streaming system. We have also collected extensive packet traces for both campus and residential access. These active and passive measurements provide insight into user behavior, traffic overhead and redundancy, and peer partnership characteristics. In the second part of the talk, we discuss a simple but insightful mathematical model for P2P live streaming. The model accounts for real-time demand for content, peer churn, heterogeneous upload capacity, peer buffering and playback delay. The model shows that performance is largely determined by a critical value. It also shows that large systems have better performance than small systems since they are more resilient to peer churn. Finally, the model shows that buffering can dramatically improve performance in the critical region.