Research On Topological Optimization Based On Interests Clustering In Unstructured P2P Networks

With the growing popularity of peer-to-peer network applications, the amount of users and data in the network are also increasing rapidly. In this case, how to quickly and accurately retrieve the resources that users need becomes an important research direction in this field.As researches show that nodes which have similar interest have a great chance to obtain the required services from each other, many scholars propose peer-to-peer topology construction strategies based on interests clustering. The existing P2P network topology construction algorithms based on interests clustering typically calculate the similarity among nodes on the base of interests extracted from the shared files. However, these strategies suffer from the following shortcomings:firstly, the free-riding nodes which account for more than seventy percent of the total number of nodes can’t join the P2P network based on interests clustering, as they don’t share any files. This is not good for the development of P2P network. Secondly, when nodes join the P2P network initially, the interests extracted from the shared files may be inconsistent with the nodes’actual interests. If nodes are clustered according to the extracted interests from shared files, they may be connected to clusters with no common interests, which obviously deviate from the original intention to build topology based on interests clustering. Finally, the strategies that use interests extracted from nodes’shared files to calculate the similarity among nodes ignore the dynamic changes of users’interests, which will lead to the problem that users can’t get the required resources effectively when nodes’ interests have changed.In order to solve the problems mentioned above, we make each node relate to two clusters of interests while optimizing the topology of P2P network in this paper. In one cluster, the node provides services to the other nodes; while in the other cluster, the node gets services from the other nodes. First of all, nodes can join the most-wanted cluster of interests at the initial stage by using "virtual interests", which can not only reduce the waste of bandwidth caused by updating topology, but also make the free-riding nodes who don’t share files accessible to the network. Secondly, after the node joined the network, we use the actual interests superimposed by retrieval interests and virtual interests to update and adjust the topology structure. In this way, we can solve the problem that nodes can’t join the most-wanted clusters precisely because of the inaccurate description of "virtual interests" when they join the network initially. Meanwhile, we can adjust the topology in time when the nodes change their interests. Thirdly, for the purpose of encouraging the nodes to share files actively and restraining the free-riding behaviors, only the nodes with similar amount of sharing files are able to connect with each other during the initial network’s construction and the subsequent adjustment. Under this circumstance, if the free-riding nodes want to get better services, they have to share more files. This strategy demonstrates that the clusters where a node gets services will be the clusters where the others provide services. In a word, a node can get the service effectively, at same time it can also provide satisfying services for other nodes.Simulation results show that the topological construction strategy improves the efficiency, and incentivize free-riding nodes to share more files.