Semantic Routing in Peer-to-Peer Systems
Currently search engines like Google, Yahoo and Excite are centralized, which means that all queries that users post are sent to some big servers (or server group) that handle them. In this way it is easy for the systems to relate IP-addresses with the queries posted from them. Clearly privacy is a problem here. Also censoring out certain information which is not 'appropriate' is simple, and shown in recent examples. To give more privacy to the users and make censoring information more difficult, Peer-to-Peer (P2P) systems are a good alternative to the centralized approach. In P2P systems the search functionality can be devided over a large group of autonomous computers (Peers), where each computer only has a very small piece of information instead of everything. Now the problem in such a distributed system is to make the search process efficient in terms of bandwith, storage, time and CPU usage. In this Ph.D. thesis, three approaches are described that try to reach goal of finding the short routes between seeker and providers with high efficiency. These routing algorithms are all applied on 'Semantic-Overlay-Networks' (SONs). In a SON, peers maintain pointers to semantically relevant peers based on content descriptions, which makes them able to choose the relevant peers for queries instead of, for example, choosing random peers. This work tries to show that decentralized search algorithms based on semantic routing are a good alternative to centralized approaches.