Lifetime characteristics measurement of a P2P streaming system: Focusing on snapshots of the overlay

Peer-to-peer streaming systems are a great success today, with millions of users. However, peer churn is still a primary challenge for the stability of P2P streaming quality, so it is essential to characterize peers' lifetime for P2P streaming systems. With a dedicated crawler for PPLive, one of the most popular P2P streaming systems, we collect substantial run-time traces, and focus our study on lifetime characteristics of coexisting peers within each snapshot of PPLive overlay, instead of analyzing whole sessions. Results obtained from above traces show that although long-lived sessions only occupy a small proportion of whole sessions, they reach an appreciable mass in any given snapshot of PPLive overlay. Inspired by this, we validate a well-known observation which can be referred to indentify long-lived peers, and point out the existing model of the observation is not accurate enough for characterizing P2P streaming systems. Furthermore, we propose a new model for above observation, and further generalize this new model to make it applicable for all channels. Evaluation results based on collected traces justify the effectiveness of the general model. Additionally, the impact of time of snapshot on residual lifetimes is also revealed in this paper.