Experiences with Complex User Profiles for Approximate P2P Community Matching

The problem of defining P2P overlays where peers characterized by similar interests are directly connected is currently an important research issue. We have recently proposed a two layer P2P architecture where the first layer exploits a gossip algorithm for the detection of communities of peers characterized by similar interests, while the second layer defines a DHT storing the profiles of the communities detected in the first layer. The DHT is exploited by peers joining the system to find out a community matching its interests. This paper investigates a DHT based approach supporting a similarity based search of user profiles. Our approach exploits Locality Sensitive Hashing to support the similarity indexing. The paper investigates several types of profiles to model user interests and evaluates the indexing mechanisms of the DHT with respect to the different types. Experimental evaluation has been conducted by considering a real data set.