Measurement Study of Shared Content and User Request Structure in Peer-to-Peer Gnutella Network

The following contribution presents the analysis of shared content distribution and user request structure in Gnutella peer-to-peer (P2P) network. Since the keyword search is the main approach to locate content in P2P systems, the shared content and user requests were studied from keywords perspective. More particularly, the keyword distributions in filenames, among peers and in user queries are studied and analytically represented. It was shown that the distribution of keywords among peers and in user queries do not follow Zipf’s distribution law, which is typical for many Internet distributions. The content of the peers was analyzed to discover similarity patterns between different peers. It has been demonstrated that all analyzed peers follow one of five main similarity patterns. In order to do so, the keyword-oriented similarity metric was proposed. In addition, the file distribution across a network was investigated and analyzed to discover replication law, willingness of peers to share files and the demographics of a network. This is intended to compare results with early presented works.