230 likes | 391 Views
MULTI-TORRENT: A PERFORMANCE STUDY. Yan Yang, Alix L.H. Chow, Leana Golubchik Internet Multimedia Lab University of Southern California. OUTLINE. Background and Introduction Proposed Approach Simulation Study Related Work and Conclusion. BIT-TORRENT (BT) OVERVIEW.
E N D
MULTI-TORRENT: A PERFORMANCE STUDY Yan Yang, Alix L.H. Chow, Leana Golubchik Internet Multimedia Lab University of Southern California
OUTLINE • Background and Introduction • Proposed Approach • Simulation Study • Related Work and Conclusion
BIT-TORRENT (BT) OVERVIEW • An extremely popular P2P file sharing application. • Nodes downloading same file (torrent) form an overlay network. • File is divided into pieces (chunks) with fixed size. • Default size is 256KB. • Upload while downloading • Nodes download chunks from their neighbors. • Each node upload file chunks to a number of neighbors. • Local scheduling file exchange.
HOW BT WORKS? • Node • Seed • Leecher • Unchoking • Tit-For-Tat (TFT) • Leecher: 4TFT + 1 optimistic • Seeds: 5 random 4TFT + 1 optimistic leecher leecher leecher leecher seed leecher 5 random seed
TORRENT IN REAL LIFE • Downloading multiple torrents simultaneously is common • Previous study shows ~85% peers participate in multiple torrents. [Guo et al. IMC 2005] • Overlap in downloading torrents: common interests do exist in real life! • Many torrents are related to each other. • Different episodes of a TV show. • Movies/music form the same genre, etc. • Start-up information are obtained from a similar source. • Similar web site, online forum, etc.
MULTI-TORRENT APPROACH • Current BT: single-torrent approach • Torrents are independent • Doesn’t take advantages of comment interest • How about multi-torrent approach? • Relating torrents together. • A natural idea but is significantly overlooked. • Encourages seeding. (Key to system performance) • Advantages of multi-torrent • Help newly joined nodes ramp up faster. • Help nodes nearing end of their downloads find the last few file chunks faster. (end-game behavior) • Keep a torrent “alive”.
CONTRIBUTIONS • A multi-torrent BT system • Small modification to the current BT protocol. • Completely decentralized, all local decisions. • A “cross-torrent” TFT (CTFT) • Simple modification to current BT TFT. • Extensive simulation-based study shows: • Multi-torrent approach improves system performance. • CTFT provides incentives for nodes to stay as seeds.
OUTLINE • Background and Introduction • Proposed Approach • Simulation Study • Related Work and Conclusion
MOTIVATIONAL EXAMPLE Staying nodes are better! Slow: +21% Fast: +17%
PROPOSED APPROACH • Current BT uses download rate for nodes to decide unchoke. Such unchoking for each torrent is independent. • Cross-Torrent TFT (CTFT) • Aggregate download rate on each torrent in peer’s all participating torrents. • Aggregation is done in a weighted manner. • Higher weight (> 1) given to the downloading rate from seeding torrents. • Normal weight (=1) given to the downloading rate from leecher torrents. • Total contribution of a peer Ny to node Nx is: Download rate on torrent i of node Nx from node Ny
OUTLINE • Background and Introduction • Motivation and Proposed Approach • Simulation Study • Related Work and Conclusion
SIMULATION SETUP • Simulator • Parameters • Event based in [Bharambe et al. INFOCOM 2006] • Extension for multi-torrent and latest BT protocol.
STAY IMPROVEMENT: # OF TORRENT Have nodes staying helps both metrics.
STAY IMPROVEMENT: INTER-ARRIVAL TIME More improvement for longer inter-arrival time. More obvious # nodes change for longer inter-arrival time.
APPLICATION: GAME PATCH SYSTEM 492MB: -77% 256MB: +84% 80MB:+92%
APPLICATION: GAME PATCH SYSTEM • Dynamic seeding (DS) • Each node does a local estimation of the ratio of seeds to leechers. • Only participate as seed in those torrents where seed to leecher ratio is below R. • This local estimation re-evaluated every S time units. 492MB: +2% 256MB: +47% 80MB: +69%
OUTLINE • Background and Introduction • Motivation and Proposed Approach • Simulation Study • Related Work and Conclusion
RELATED WORK • Explore multi-torrent • Multi-torrent system using tracker overlay. [Guo et al. IMC 2005] • Our focus is on performance. • Our CTFT provides incentives to stay as seeds. • Our design requires only local client modification. • Explore BT incentives in single torrent • A market based incentives. [Freedman et al. IPTPS 2008] • Reputation based incentives. [Lian et al. IPTPS 2006] • Our CTFT-DS matches seeding capacity demand and supply with local knowledge. • Investigating other incentive approaches are part of our future work.
CONCLUSION • Performance gains are possible through multi-torrent. • Nodes staying as seeds improve performance. • CTFT provides incentive for nodes to stay as seed. • CTFT-DS makes seeding efficient. • Multi-torrent is a promising research area with remaining future directions. • Malicious users. • More exploration on system parameters. • R, S, CTFT weight, etc. • Real word implementations.
Q & A • Thank You!
STAY IMPROVEMENT: CTFT WEIGHT Staying nodes perform better in both class.
STAY IMPROVEMENT: FRACTION OF STAY Incentives! No Incentives
DISCUSSION (BACKUP) • Performance in the real world • We expect even bigger improvement. • System parameters • Exploration of various system and proposed schemes’ parameters is our on-going effort. • Malicious behavior • A malicious node can take advantage of the weight of CTFT, how to detect malicious node is one of our future works. • How malicious behavior affects the system performance is one of our future works.