Measurement and Analysis on P2P Swarms Based on Coupon Collector's Problem

Abstract

Active measurement on peer-to-peer system is the best way to understand the current situation and its evolution trends. And it also provides the real measurement basis to support the research on system simulation and optimization. Based on coupon collector's problem in random sampling, rapid swarm covering model is proposed to guide the active measurement on BitTorrent swarm, which can predict the number of measurement requests is fast convergence in the mathematical expectation. And with the optimization of lowering expectations and requesting concurrently, the request repetition rate is 67.12% off, while still covering the 95% peers in a swarm. Moreover, the measurement time can be further shortened by interacting with peers by DHT and PEX. Based on the theoretical model, a low cost measurement platform is set up to trace the 49,854 Torrentz swarms and 98 HDChina swarms for a long time. Furthermore, the geographical characteristics are also analyzed. Index Terms - P2P swarm; active measurement; coupon collector's problem; geographical characteristics; user preference