| As a new network compute mode , P2P (peer-to-peer ) network modifies the traditional C/S mode and is used more and more popularity. With the increasing appilication of P2P network , distributed search technology has been one of the most important issues in the area of search. Recently , the search mechanism of existed structured or unstructured P2P systems , both have advantages and disadvantages. Search mechanisms ( like flooding )in unstructured P2P networks are inherently blind , which makes the search inefficient and unscalable, and with the scale of the P2P network growing, it generates exponentially increasing redundant messages .While DHT_based structured P2P networks can provide search effiency and scalability by deploying identifier-based search mechanisim ,they fail to support flexible multi-keyword search like unstructured P2Pnetworks do.In the thesis , we analysed the search mechanism of unstructured P2P networks . In order to solve the problem in unstructured P2P networks ,such as inherently blind,search inefficient and redundant messages , we proposed a modified interest-domain based search model .In this method , every node in the unstructured P2P network keeps LRK (Local Resource Keywords) ,LQK( Local Query Keywords) and Friend List. Comparing the similarity of the keyword in nodes'LRK and LQK , peer put the other peers who have the same interest into its Friend List. This will give the direct for the next step. Search requests are propagated within the interest-domain , it will ask the node in the Friend List first , so , it will get the related information quickly , improve the search efficiently and result in less messages. With more and more search , by learning other peers'interests, peers can get more relevant resources faster , reduce the search time with less message traffic.Model analysis and simulation prove that, the interest-based search mechanism is more efficient than the traditional search method in Gnutella-like system. |
PDF Full Text Request