Research On Effective Search Mechanisms In Peer-to-Peer Networks

With the rapid growth of Internet and computing power, peer-to-peer (P2P) networks have gained much attention from both industrial and academic fields. P2P networks share idle CPU power, free diskspace and network bandwidth between different nodes in a distributed and equal way. As for any heavily used large distributed systems, the effectiveness of a P2P network largely depends on the versatility and scalability of its search mechanism.Besides centralized systems, P2P networks can be roughly classified into two categories: unstructured and DHT-based structured P2P networks. Search mechamisms in unstructed P2P networks are inherently blind, which makes the search inefficient and unscalable. While structured P2P networks can provide search efficiency and scalability by deploying identifier-based search mechanism, they fail to support flexible multi-keyword search like unstructured P2P networks do.In the thesis, an interest-domain based search model in unstructured P2P networks is proposed. Documents are described by metadata using RDF. Interest-domain is formed by nodes sharing same metadata. Search requests are propagated within the interest-domain and thus efficient. Metadata selection window and metadata replication can further improve the search efficiency. Model analysis and simulations prove the effictiveness of the proposed search mechanism.After that, a multi-keyword search model based on keyword association in structured P2P networks is proposed. The search model incurs much less network traffic than the naive multi-keyword search model, which is proved by simulations. Another multi-keyword search model based on metadata complete query expression is proposed, which can finish a multi-keyword search by contacting just one node at the cost of diskspace. To reduce the cost, a load balacing mechanism is also proposed.Finally, a semantic-based search model over arbitrary topology in P2P networks is proposed in the thesis. By separating semantic search layer from P2P layer, the proposed model can provide very flexibile search functions. Semantic descriptor is used to perform query matching and routing. Through the mechanisms of query history and caching of semantic descriptors along the search path, the performance of the search model is highly improved, which is also showed by simulations.