Node Selection Algorithm Analysis Of P4p Based On Similarity Theory

With the development of the Internet, there are more and more network users using P2P software, which shares resources more easily and improves the online transferring rates and so on. However, using P2P software in large-scale has also brought enormous pressure to the broadband network. According to the current survey, the P2P traffic has exceeded the HTTP traffic and has become the main character of the bandwidth network. This now also becomes a headache for the network operator. To fix this problem, network operators have applied a variety of methods to monitor P2P users and to restrict traffic, but no efforts can solve the current problems between network operators and P2P users.In fact, the fundamental reason of this problem is not for the application itself, but rather the exchange mechanism of P2P network. Since the network emphasizes too much on the network between nodes "reciprocity", while the node being mapped to cover the line, making the physical nodes and similar Internet are not necessarily similar in coverage. Therefore, it is possible to download resources from a host far away, wasting network bandwidthFor the reasons mentioned above, YONGHAIXIE etc. from Yale University presented P4P architecture. P4P means Proactive Network Provider Participation for P2P [5] or the Provider Portal for Application [6], which is the network providing benefit for the P2P network provider or agreement in the application of network service. Through P4P system, we can introduce the server interface between the network layer and the P2P application layer to assist the P2P application for searching nodes and node selection, thereby reducing the pressure on backbone and ISP bandwidth.In this research, a node selection algorithm is proposed under the P4P system, based on the principle of similarity. In the analysis of the research background and full introduce of the basic structure of the P4P system, the essay using the similar principle carry on the cluster according to the geographical position closely, Then using the binary tree thought to carry on the division to the cluster node, as far as possible to make each PID in the network achieve balanced. Designed two of the p-distance, to prove minimize the network backbone traffic MLU and the implementation mechanism. ISP saved PID two-dimensional table in the iTracker, guiding the network nodes in queries globally, updating the network information can be completed at the same time.In addition, the paper proposed the algorithm for the simulation experiment, proved that the network efficiency of P2P users have a certain increase under the system.