Research On Interest Domain Based Search Schemes For Unstructured Peer-to-peer Network

The great development and growing popularity of Internet lead to the explosive increment of Internet resources. To help people share and download resources from the Internet conveniently, more and more peer-to-peer (P2P) softwares have been developed. There has been a huge success of P2P software which thoroughly breaks the traditional Client/Server mode and puts forward an equal member mode. In P2P network, every peer is equal, both suppliers and consumers of resources are based on the idea of"All for one, one for all". The most attractive and popular P2P software is the unstructured P2P software. The unstructured P2P network has the most pure peer-to-peer distributed network architecture, and it owns several characteristics including "variety", "random access", "self- adaptive" and "growth",etc. Large numbers of resources and documents decentralize on the distributed numerous nodes and they are independent of any administrative node or central node, node freely share and download resources, and also freely join and leave the network without influencing the whole network, therefore the unstructured P2P network is the most suitable system for sharing and downloading resources in current network.Because the unstructured P2P network has the characteristics of random access, indeterminate and distributed structure, the key problem of this network search scheme is how to search resources efficiently. So far, most existing search schemes for the unstructured P2P network use flooding search schemes, or modified flooding search schemes. Along with the aggrandizement of network scale and the increase of network nodes, the redundant search messages will increase in an exponential way, which would seriously influence the efficiency of the unstructured P2P network search schemes. Meanwhile, the research on"small-world"model and the"power-low"distribute theory show that: the structure of the unstructured P2P network is not a completely random structure, and it has the characteristics of both"small-world"and"power-low"structure. The nodes are self-moving to form an interest domain, the inter-domains existing short conjunction chains, and in the intra-domain, only the minorities of nodes have higher degrees than the others.Motivated by the topology of the unstructured P2P network and the analysis of existing search schemes for unstructured P2P network, in this thesis, we propose an interest-based search scheme for unstructured peer-to-peer network. In this search scheme, we use a budget value and TTL (Time To Live) value to control the number of searched nodes and the search depth. We divide the nodes of the Interest-domain network into three types: common nodes in the intra–interest-domain, near-partition-nodes which connect an interest domain edge and partition-nodes which are in the inter-interest-domain. Besides, we classify the search document types of the Interest-domain network into four types: fashion document inside the interest domain, non-fashion document inside the interest domain, fashion document outside the interest domain, and non-fashion document outside the interest domain. Thus, for different types of nodes, using different budget assignment strategies to search different types of files helps to improve the efficiency of message search, reduce the numbers of redundant search nodes, and improves the efficiency of the search scheme as a result.The last part of the thesis is the analysis and simulation of the proposed search schemes. Comparing with the representative search schemes for the unstructured P2P network, by using our proposed search scheme, a lot of redundant messages have been avoided and the utilization rate of query messages has been significantly improved; at the same time, success rate and efficiency of the search scheme have also been improved. These advantages show that considering budget control and different search schemes for different topologies and different search types have great practical value in P2P network.