A substantial amount of work has recently gone into localizing BitTorrenttraffic within an ISP in order to avoid excessive and often times unnecessarytransit costs. Several architectures and systems have been proposed and theinitial results from specific ISPs and a few torrents have been encouraging. Inthis work we attempt to deepen and scale our understanding of locality and itspotential. Looking at specific ISPs, we consider tens of thousands ofconcurrent torrents, and thus capture ISP-wide implications that cannot beappreciated by looking at only a handful of torrents. Secondly, we go beyondindividual case studies and present results for the top 100 ISPs in terms ofnumber of users represented in our dataset of up to 40K torrents involving morethan 3.9M concurrent peers and more than 20M in the course of a day spread in11K ASes. We develop scalable methodologies that permit us to process this hugedataset and answer questions such as: "\emph{what is the minimum and themaximum transit traffic reduction across hundreds of ISPs?}", "\emph{what arethe win-win boundaries for ISPs and their users?}", "\emph{what is the maximumamount of transit traffic that can be localized without requiring fine-grainedcontrol of inter-AS overlay connections?}", "\emph{what is the impact totransit traffic from upgrades of residential broadband speeds?}".
展开▼