Understanding BitTorrent: A reality check from the ISP’s perspective

During the last decade, Peer-to-Peer (P2P) applications have been one of the dominant components of Internet traffic. Understanding BitTorrent, by far the most popular P2P protocol for data distribution, is extremely valuable to shed some light on the nature of distributed systems. This paper surveys the existing measurement studies and sets out to verify the acquired corpus of knowledge about BitTorrent by analyzing the largest and most comprehensive data-set so far. We collected BitTorrent traffic at four major European ISPs during 2009 and 2010, a vantage point not yet exploited by previous measurement studies. Our analysis puts into perspective and corroborates several well-known findings, such as that: (1) 20% of the most popular torrents represent more than 95% of the BitTorrent activity, (2) only 1–3% of the BitTorrent traffic stays local, i.e., within an ISP, (3) 4–44% of the BitTorrent traffic could be localized using appropriate locality-awareness techniques, and (4) about 20% of downloads get stalled due to scarcity of content pieces.