Boxun Zhang: An Empirical Approach to Understand BitTorrent

BitTorrent has been one of the most popular Peer-to-Peer file sharing applications in the last decade, and it generates significant amount of Internet traffic. Although the protocols and algorithms used by BitTorrent
are simple, it is not trivial to understand how BitTorrent is operated and how BitTorrent users behave in real world, and such knowledge is key for researchers to improve BitTorrent and build new applications like video streaming, file backup, and social networks upon it.

To obtain such knowledge, we have performed measurements of many different BitTorrent communities, and we turn those effects into the P2P Trace Archive, a place for researchers to exchange traces. We also look into the strength and limits of various measurement techniques, and try to understand how those techniques affect the measurement results. Using our recent datasets from two of the largest BitTorrent trackers, we investigate the flash crowd phenomenon in BitTorrent, and reveal many interesting facts that are previously unknown to the community. Interestingly, we find that BitTorrent does not always perform well in flash crowds, which differs from the impressions of many of us.

The speaker was invited by Gunnar Kreitz.