Multimedia Access Community for Unstructured P2P Networks

Peer-To-Peer (P2P) network users expect to be able to search objects by semantic attributes based on their preferences for multimedia content. Partial match search has become an essential service in P2P systems. To realize an interest-based small-world CommunityNet, we also investigate the following practical design issues: 1) similarity estimation: we define a quantifiable similarity measure that enables clustering of similar peers in CommunityNet; 2)
distributed small-world overlay adaptation: peers maintain a smallworld overlay under network dynamics; and 3) query strategy under the small-world overlay: we analyze appropriate settings for the Time-to-Live (TTL) value, for TTL-limited flooding, that provides a satisfactory success ratio and avoids redundant message overhead. We use simulations and a real database called AudioScrobbler [1], which tracks users’ listening habits, to evaluate the performance of
CommunityNet. The results show that CommunityNet assists peers in locating content at peers with similar interests through short path lengths, and hence, achieves a higher success ratio (than nonsmallworld interest-based overlays and noninterest-based small-world overlays) while reducing message overhead significantly.