P2PQueryStrings.PDFVIP

Geographical Statistics and Characteristics of P2P Query Strings Adam Shaked Gish Yuval Shavitt Tomer Tankel P2P Networking Research, School of Electrical Engineering, Skyrider, Inc., Israel Tel-Aviv University, Israel Email: adams@ Email: {shavitt,tankel}@eng.tau.ac.il Abstract— P2P file-sharing applications are quickly advertisers will be able to expose them to relevant being adopted by a wider and more mainstream targeted advertising content. In order to improve audience. There is much to be learned from keyword advertising relevance, it is critical to analyze the searches users perform in order to retrieve content behavior and preferences of P2P users, taking into from these networks. account the user’s geographic location and temporal This paper presents a large-scale measurement behavior. study of search terms in the modern Gnutella net- work. We developed a highly parallelized architecture Most measurement studies on P2P content have capable of capturing an unprecedented amount of focused either on analyzing ISP data traffic [1], [2] geographical-identified queries. We applied this ar- or the content in user’s shared folders [3], [4], [5], chitecture to generate daily logs of search queries. [6]. The popularity of files [3], [5], [6] follows a We collected over 25 such daily Gnutella logs, with more t