Clustering in P2P Exchanges and Consequences on Performances

We propose here an analysis of a rich dataset which gives an exhaustive and dynamic view of the exchanges processed in a running eDonkey system. We focus on correlation in term of data exchanged by peers having provided or queried at least one data in common. We introduce a method to capture these correlations (namely the data clustering), and study it in detail. We then use it to propose a very simple and efficient way to group data into clusters and show the impact of this underlying structure on search in typical P2P systems. Finally, we use these results to evaluate the relevance and limitations of a model proposed in a previous publication. We indicate some realistic values for the parameters of this model, and discuss some possible improvements.