Peer-to-Peer System Design

Abstract

It has gradually become clear that Peer-to-Peer (P2P) systems should not be conceived in the manner of conventional computing systems. Consequently, over the years, ideas from social science in general and economics in particular have made their way to P2P systems to deal with the novel challenges that such systems throw up. In Chapter 2 of this thesis we examine many of the socioeconomic ideas that have been employed in the literature. We discover that the economic ideas that have been utilized for incentive related works in P2P systems are imbued in the rational action framework. The assumptions entailed in this framework drive the solutions to free-riding in P2P incentive works, and also impinge upon the usage of terms such as fairness, social welfare, etc. Our examination leads us to accomplish three major tasks: 1) We argue for the reconsideration of commonly used concepts and paradigms; 2) We propose novel socioeconomic concepts and approaches for the design of P2P systems; and 3) We open up new areas that could benefit from the use of socioeconomic ideas. In Chapter 3 we take inspiration from opinion gathering techniques such as VoxPopuli; voting tools such as the Ballot Box; and general acceptance of the opinion of \emph{experienced} members of the society, to devise a voting infrastructure to fight spam in Tribler, a P2P media distribution system In Chapter 4 we indicate that private BitTorrent communities that use sharing ratio enforcement for incentivizing contribution from peers, can face problems such as credit crunches. In Chapter 5, we show that such problems are exacerbated when peers do not follow the rational user model. We also present some ways in which these problems can be rectified. In Chapter 6 and 7, we argue for the consideration of alternate viewpoints viz a viz concepts such as fairness, freeriding and social welfare; and for formulation of alternatives to game-theoretic solution concepts for assessing the robustness and performance of distributed protocols. In Chapter 6 we propose effort based incentives with a novel concept of fairness, for improving efficiency and fairness in P2P systems. In Chapter 7 we argue that the goal of most earlier P2P works, of achieving a Nash equilibrium when all peers are rational, is limited. We present Design Space Analysis, a simulation based method that: allows for flexible user behavior assumptions; aims to analyze protocols more comprehensively and not just on a single point in the design space; and provides a tournament based approach for evaluating protocol robustness and performance.