A task-based model for the lifespan of peer-to-peer swarms

Abstract

Peer-to-Peer (P2P) techniques are broadly adopted in modern applications such as Xunlei and Private Tracker [1,2]. To address the problem of service availability, techniques such as bundling and implicit uploading are suggested to increase the swarm lifespan, i.e., the duration between the birth and the death of a swarm, by motivating or even forcing peers to make more contributions. In these systems, it is common for a peer to join a swarm repeatedly, which can introduce substantial bias for lifespan modeling and prediction. In this paper, we present a mathematical model to study the lifespan of a P2P swarming system in the presence of multi-participation. We perform evaluations on three traces and a well-known simulator. The result demonstrates that our model is more accurate than previous ones. © 2012 IFIP International Federation for Information Processing.