Modeling and performance analysis of BitTorrent-like peer-to-peer networks

Abstract

In this paper, we develop simple models to study the performance of BitTorrent, a second generation peerto-peer (P2P) application. We first present a simple fluid model and study the scalability, performance and efficiency of such a file-sharing mechanism. We then consider the built-in incentive mechanism of BitTorrent and study its effect on network performance. We also provide numerical results based on both simulations and real traces obtained from the Internet.

A mathematical model for file fragment diffusion and a neural predictor to manage priority queues over BitTorrent

“…For networks consisting of a large number of nodes, some priority management systems are based on scalable algorithms that ensure rapid convergence, such as Epidemic-style or gossip-based algorithms as in the work of Ghit et al (2010). Another approach is that of Qiu and Srikant (2004), who propose to rank peers according to their upload bandwidth; hence, when having to provide some contents, the selection of peers is performed accordingly. One of the mechanisms proposed to increase file availability has been to use a multi-torrent, i.e., for ensuring fairness, instead of forcing users to stay longer, contribution is provided to uploaders for fragments belonging to different files (Guo et al, 2005).…”

Correlating Topology and Path Characteristics of Overlay Networks and the Internet

“…This mechanism ensures that for a user the amount of incoming data is roughly equal to the amount of outgoing data. The BitTorrent bartering-partners selection protocol aims at filling up all the available bandwidth of a peer; for popular files, the bandwidth filling property is guaranteed [19]. For this reason, and because of the lack of search features at the peer level, BitTorrent is bandwidth-limited.…”

“…After sorting the shared files according to their swarm size, the measurements setup module, selects the top-N files from the largest web site, and their respective torrents are downloaded. The number of files measurement parameter (N) reflects the physical resources in the measurement infrastructure: the more measurement machines, the larger the possible value of N. In practice, the maximum value of N also depends on the number of files with swarms of sizes of a few tens of peers, since only these files can actually be reliably and efficiently obtained from the network [19].…”

Optimal policies for control of peers in online multimedia services

“…As a byproduct, the analysis also shows in the limit the switching curve is independent of model input parameters. Qiu and Srikant [3] use fluid and diffusion approximations to model the behavior of P2P networks since it is not tractable to model it using standard queueing methods even if the system has Markovian properties. Usually, these approximations scale both in time and state space, and get a limit process that is mathematically tractable using functional strong law of large numbers (FSLLN) and functional central limit theorem (FCLT); for the details of FSLLN and FCLT, see Whitt [11].…”

“…Most of the recent research on P2P networks focuses on modeling and performance analysis (Ge et al [1], Clévenot and Nain [2], Qiu and Srikant [3]), or optimal peer search and selection (Adler et al [4]). However control of P2P networks is relatively unexplored.…”