A case for coalitions in data swarming systems

Abstract

We present an argument in favor of forming coalitions of peers in a data swarming system consisting of peers with different upload capacities. A coalition is a set of peers with the same upload capacity that explicitly cooperate with other peers inside the coalition via choking and capacity allocation strategies. Further, each peer interacts with other peers outside its coalition via potentially distinct choking and capacity allocation strategies. This paper focuses on the efficiency of different choking strategies, assuming that peers do not share data with other peers outside their coalitions. We first develop an analytical model that accurately predicts the performance of a coalition of peers adopting BitTorrent's Tit-for-Tat choking strategy. Our model highlights a number of inefficiencies of Tit-for-Tat strategy. Accordingly, we propose a random choking strategy, and show that it can help a coalition achieve near-optimal performance and it significantly outperforms not only Tit-for-Tat strategy but also unchoke-all strategy. Using cooperative game theory, we prove the existence of stable coalitions, and demonstrate the convergence of the dynamic coalition formation process when peers use our cooperation-aware better response strategy. Using extensive simulations, we demonstrate significant performance benefits due to coalition formation.