Zipf-Law: A Measurement Study Of Geographical Distribution Of Peers In Bittorrent Swarms

In the past few years, many locality-based approaches have been proposed to reduce cross-ISPs traffic caused by BitTorrent (BT). However, most of these approaches have taken such an assumption that there are always plenty of peers in client's neighborhood, which results in inefficiently reducing cross-ISPs traffic. In this paper, we study the characteristics of geographical distribution of peers in BT swarms. By measuring and analyzing more than 2 000. torrent file swarms of BT in 5-month, the geographical distributions of peers is found to follow Zipf-like law both in country level and AS level. More specifically, the distributions of peers exhibit a non-Zipf head, but Zipf tail. Our work shows that the characteristics of geographical distributions of peers can be used to better understand and answer questions related to traffic optimization of BT.