Reverse-engineering BitTorrent: A Markov approximation perspective

In this paper we understand BitTorrent protocol from a Markov approximation perspective. We show that together with the underlying rate control algorithm, the rarest first algorithm and choking algorithm in BitTorrent protocol implicitly solve a cooperative combinatorial network utility maximization problem in a distributed manner. This understanding allows us to access properties of BitTorrent from a fresh perspective, including performance optimality, convergence and impacts of design parameters. Our numerical evaluations validate the analytical results.