A SIMULATIVE FRAMEWORK TO ESTIMATE P2P PERFORMANCE INDEXES AND TRAFFIC MATRICES

Peer-to-Peer (P2P) applications today cause about 60-80% of the overall Internet traffic. P2P traffic stresses network infrastructures with highly dynamic and distributed traffic matrices, whose vertices heavily depend on the overlay network and on the query mechanism adopted. Two of the most important characteristics of real P2P systems are the huge number of peers and their very dynamic behavior. In this respect, our aim is to propose a simulation approach capable to represent the peers' and files' dynamics in very scalable way. This approach is able to give indications about the file distribution behavior and queries' success percentage along time and to support network managers or traffic modeling tools in determining the impact of P2P applications on transport networks. Simulation results are reported which show the ability of the system of representing overlay networks of about 250000 peers.