Evaluating stranger policies in P2P file-sharing systems with reciprocity mechanisms

How to deal with strangers efficiently is a big challenge for designing P2P systems, as ineffective stranger policies will cause the degradation of system performance and fairness. In this work, we focus on an open P2P file-sharing environment with reciprocity incentive mechanisms, and we evaluate the impact of different stranger policies to system performance and fairness using both numerical analyses and agent-based simulations. The results reveal that there exist tradeoffs between system performance and fairness; the optimal performance and fairness of the system cannot be reached simultaneously; free-riders will survive when system performance is optimal. The accuracy of this evaluation is verified through the analysis of stranger policies in BitTorrent and eMule/eDonkey. Therefore, this evaluation can effectively help P2P designers select appropriate stranger policies according to their individual design goals.