Evolution of Resource Reciprocation Strategies in P2P Networks

In this paper, we consider the resource reciprocation among self-interested peers in peer-to-peer (P2P) networks, which is modeled as a stochastic game. Peers play the game by determining their optimal strategies for resource distributions using a Markov decision process (MDP) framework. The optimal strategies enable the peers to maximize their long-term utility. Unlike in conventional MDP frameworks, we consider heterogeneous peers that have different and limited ability to characterize their resource reciprocation with other peers. This is due to the large complexity requirements associated with their decision making processes. We analytically investigate these tradeoffs and show how to determine the optimal number of state descriptions, which maximizes each peer's average cumulative download rates given a limited time for computing the optimal strategies. We also investigate how the resource reciprocation evolves over time as peers adapt their reciprocation strategies by changing the number of state descriptions. Then, we study how resulting download rates affect their performance as well as that of the other peers with which they interact. Our simulation results quantify the tradeoffs between the number of state descriptions and the resulting utility. We also show that evolving resource reciprocation can improve the performance of peers which are simultaneously refining their state descriptions.