Churn: a Key Effect on Real-World P2P Software

Abstract

Churn refers to a large number of arriving and departing participants within a short time in peer-to-peer (P2P) networks. This paper studies the impact of churn on real-world unstructured P2P networks. To this end, we need collecting snapshots of P2P network topology and associated performance metrics. Because P2P topology changes dynamically, the time to take a snapshot must be sufficiently short for the snapshots to be accurate. We propose Third-party-to-servent Crawling with Servent-to-servent Sampling (TCSS) system. TCSS uses a third-party crawling technique to collect network topology information without disturbing the original P2P network under investigation. Furthermore, TCSS adopts distributed and parallel techniques to speed up the crawling process. TCSS also employs a servent-to-servent sampling technique to gather the corresponding performance metrics of the P2P network simultaneously. Empirical results show that TCSS takes around 7 minutes to take a topology snapshot of the P2P network. Besides, we found that churn is indeed a combined effect of peer arrivals/departures and neighbor replacements. As the number of peers increases, the number of very long-lived peers remains nearly constant and the P2P network possesses a small-world property. Moreover, as churn aggravates, the average booting time of peers increases and the variation is proportional to the degree of churn. The response time of the top-rank keyword searches is not affected by the degree of churn.