4.1.5 Similarity based Searching
Consider a peer1 generates the query request for an object O and then peer5 hosting objects O. Instead of blindly flooding query messages in the typical Gnutella protocol, the search protocol conceptually operates as follows. Among the neighboring peers in Ipeer2, peer1 selects the peer2 which have high similarity value than peer3 (peer2,Q) > (peer3,Q).Upon receiving query Q, peer 3 performs similarly by selecting its neighbors who are more similar to peer5 than peer3 hoping that there is at least one overlay path P toward the destination pee 5, such that any two neighboring peers Peer4 and peer5 on P (peer5 receives Q sent from peer3 ) have(peer5, Query) > (peer4,Query).Clearly, the path P intends to explore progressively the peers that are likely to resolve the query Q.
Fig. 2 Similarity Computation
V. PROPOSED SYSTEM
This project proposes a novel decentralized overlay construction algorithm provides rigorous performance guarantees to enhance the efficiency and effectiveness of searches in unstructured P2P networks. In unstructured P2P network G = (V, E), where V is the set of participating peers, and E is the set of overlay connections linking the peers in V.
The peers in G may be interconnected randomly therefore the main aspect of this project is to restructure and optimize the unstructured p2p overlay based on following properties such as high clustering among similar peers, low diameter among the distinct peers to rapidly propagate the query message between them and overlay links must be very progressive. This approach constructs the optimized overlay in a decentralized manner by exploiting the similarity of participating peers. Therefore, it does not need centralized servers to perform query searching. In network formation algorithm, a querying peer takes hop count (where c is a small constant) to reach the destination peer by progressively and effectively exploiting the similarity of peers on the search path. Additionally, the probability of successfully discovering a requested object in this approach is approximately 100 percent. This approach exhibits file sharing pattern based on power-law property that leads to achieve rigorous performance guarantees in terms of the hop count of routing a query message and the successful query ratio. In proposed approach, network host the objects of interest and maintain no foreign indices that eliminates storage and bandwidth overheads for publishing and managing such indices.
Fig. 3 Proposed Architecture
The main motivation of this scheme is improve the overall performance of P2P searching scheme. The main advantages of this approach are High quality results, Low search cost, Incurs less hop count...