Internet traffic continues to exhibit exponential growth in the past few years. This forces Internet service providers(ISPs) to continuously invest in infrastructure upgrades and deploy traffic management techniques, such as caching and locality, to fulfill the increasing user demand. To help ISPs better manage their infrastructures, it is important to compare and understand the similarity of their user interests. However, such a comparison is challenging because the ISP data is hard to obtain, not to mention the related modeling and analysis issues. In this thesis, we aim to understand the ISP similarity through an extensive analysis of Peer-to-Peer(P2P) user interest. To collect the P2P dataset, we develop a tool to automatically download BitTorrent's meta-info(torrent) files on the Internet. This tool also helps us to collect important peer and content information in these BitTorrent swarms without uploading any copyrighted files. As a result, we successfully obtained 16,697 active peers from 1,721 torrents in 1,097 unique Autonomous Systems(ASes). After that, we adopt the classic statistical and clustering approaches to compare their different user interests. Our research for the first time shows the existence of cloud users in such real-world content distribution systems as BitTorrent. The model analysis further indicates that we can adopt similar traffic management approaches (e.g., caching similar contents) across geographically closer ASes.
University of Minnesota M.S. thesis. June 2019. Major: Computer Science. Advisor: Haiyang Wang. 1 computer file (PDF); 63 pages.
Understand the Similarity of Internet Service Providers via Peer-to-Peer User Interest Analysis.
Retrieved from the University of Minnesota Digital Conservancy,
Content distributed via the University of Minnesota's Digital Conservancy may be subject to additional license and use restrictions applied by the depositor.