Rarest first and choke algorithms are enough
Proceedings of the 6th ACM SIGCOMM conference on Internet measurement
Lightweight emulation to study peer-to-peer systems
Concurrency and Computation: Practice & Experience - Hot Topics in Peer-to-Peer Systems (HoTP2P2006)
Bittorrent is an auction: analyzing and improving bittorrent's incentives
Proceedings of the ACM SIGCOMM 2008 conference on Data communication
Taming the torrent: a practical approach to reducing cross-isp traffic in peer-to-peer systems
Proceedings of the ACM SIGCOMM 2008 conference on Data communication
A BitTorrent Performance Evaluation Framework
ICNS '09 Proceedings of the 2009 Fifth International Conference on Networking and Services
Do incentives build robustness in bit torrent
NSDI'07 Proceedings of the 4th USENIX conference on Networked systems design & implementation
The bittorrent p2p file-sharing system: measurements and analysis
IPTPS'05 Proceedings of the 4th international conference on Peer-to-Peer Systems
A performance study of BitTorrent-like peer-to-peer systems
IEEE Journal on Selected Areas in Communications
Characterizing the file hosting ecosystem: A view from the edge
Performance Evaluation
Node coordination in peer-to-peer networks
COORDINATION'12 Proceedings of the 14th international conference on Coordination Models and Languages
| Hi-index | 0.00 |
BitTorrent (BT) is currently the main P2P protocol used for sharing large files over the Internet. It is therefore important to understand the performance characteristics of existing real-world BT implementations (clients). In this work, we measure the download speed of over 10 million BT users over one month. Surprisingly, our measurements show that the two most famous BT clients, namely uTorrent and Vuze (Azureus), achieve different download speeds for the same set of torrents. In particular, we observe that uTorrent users achieve 16% faster download speed than users of Vuze in our data set. To shed light to the cause of this difference, we reverse engineer the two clients to infer their individual design choices. Our study shows that the two clients differ in two important areas: (a) how they manage their neighborhood, and (b) how they distribute their upload capacity to their peers. We speculate that the cause of the mismatch in download speeds can be attribute to these differences. We hope that our findings will open the door for new research efforts to better understand the impact of design choices in the performance of real-world BT implementations.