An Unstructured P2P Resource Search Strategy Based On Semantic Clustering

Resource search is widely discussed in the research area of P2P system, unstructured P2P resource search mainly uses flooding mechanism to solve the problem, however, hardly can querying efficiency be guaranteed with the increase of retransmit of query request, more network jams and severe waste of bandwidth. As a result, resource search is still a challenge that large-scale unstructured P2P network should be confronted with.Accordingly, this paper advances a strategy of resource search based on semantic clustering which we name it LSCS(Local semantic clustering search). The follow writings will clarify some main work and some original ideas of this paper.Firstly, with respect to network structure, this paper builds a double-layer distributed structure, advantages of which are no center servers, peering relationships between nodes, as a result, problems such as performance bottle-neck and failure of single point will not appear.Secondly, with regard to resource expression and local clustering, this paper improves the vector space model (VSM) since many ways of resource expression based on DHT can not accurately express the semantics of resources. Changing the resource vector to semantic vector avoids the inaccuracy since we assume the similarity between characteristic values is equivalent to that between semantics. Moreover, for it is hard to cluster between high dimensional vector, this paper proposes an algorithm of construct semantic forest (CSF) to cluster local resource, construct semantic node layer. Since no information exchange between nodes happens, no waste of network bandwidth exists, as a result, the system has better expansibility.Thirdly, in regard to resource search, according to the blindness of retransfer of query request in unstructured P2P system, this paper establishes similar link between nodes, so as to cluster the nodes with similar resources. As a result, we get the set of similar semantic nodes. Furthermore, optimizing these similar links dynamically makes the sets nearer in physical network. More importantly, this paper defines the combination of orientation mode and flooding, mode, adds a dynamic TTL mechanism, so that to increase the efficiency of resource search.Finally, the analysis of performance and experimental tests represent that the strategy advanced in this paper is practicable. It can raise the search efficiency of the system, moreover, it can increase the rate of successful search.