Exploring the sustainability of credit-incentivized peer-to-peer content distribution

Abstract

Credit-based incentives were proposed to incite peer contributions in P2P content distribution systems. Their effectiveness was extensively analyzed from a game theory perspective. Little attention however has been paid to a potential threat to such systems - the possible condensation of credits in a small number of peers over time. Credits condensation puts system sustainability on the line: many peers gradually run out of credits and cannot keep up a decent download rate. We study the sustainability of credit-based P2P systems running for a long period of time. We first introduce a new queueing network based model for credit circulation in a P2P content trading market. This model enables the study of credit system sustainability via examining the stability of stochastic traffic flows in a network of queues. We show that a stable job circulation, i.e., an equilibrium market state, always exists. A sufficient and necessary condition for asymptotic condensation at equilibrium is proved. We analyze the degree of condensation in finite networks using the Gini index, and relate condensation to P2P network protocols and parameters. Our theoretical results are verified and supported by extensive simulations under realistic settings. We propose counter-actions for preventing and mitigating credit condensation. © 2012 IEEE.