FAIR TORRENT: A DEFICITBASED DISTRIBUTED ALGORITHM TO ENSURE FAIRNESS IN PEER-TO-PEER SYSTEMS

By Chaithra Iyr Mtech CNE

CONTENTS
INTRODUCTION BACKGROUND AND RELATED WORK FAIRTORRENT ALGORITHM FAIRTORRENT PROPERTIES EXPERIMENTAL RESULTS CONCLUSION REFERENCES

INTRODUCTION
Problem
P EER-TO-PEER (P2P) file sharing is a popular, cheap, and

effective way to distribute content. However, P2P filesharing applications suffer from a fundamental problem of unfairness.
Free-Riders and Low-Contributing peers Consume much bandwidth in P2P networks Cause much slower downloads for other users High-Contributing peers often receives much less bandwidth

than they contribute.

Tit for Tat (TFT) BitTyrant Block-based TFT PropShare Proportional Response algorithm. Above approaches are rate based and suffer from fundamental flaw. Fair Torrent(FT): a new deficit-based distributed P2P algorithm that solves the problem of fair bandwidth exchange in the presence of free-riders and strategic peers. FairTorrent accurately rewards peers in accordance with their contribution. FT runs locally at each peer and maintains a deficit counter for each neighbor.
Bit Torrent

 Advantages Of FT

-Fair bandwidth allocation

-resilient to free riders

-Maximize upload capacity utilization -avoids long peer discovery & reaches a fast rate convergence. -no need to estimate and predict peers allocation

-A FT client requires no changes to Bittorrent protocol

 Fair torrent is implemented inside a Bittorrent client and

its fairness and performance is compared against 4 other widely used open source implementation. ie
1.

The original BitTorrent Python client by Bram Cohen ,

2. The popular open-source Azureus Java BitTorrent client

3. The PropShare client 4. The strategic BitTyrant client.

Background And Related Work

BitTorrent Overview Seed File: Seed

Leechers

BitTorrents Tit-for-Tat (TFT)

Operates by having each BitTorrent client upload to N other peers in a round-robin fashion, where N-k of the peers have provided the best download rate during the most recent 1020-s. Choking and un choking Unstable peer relationships

2.5

1 0.5 1

Peer i
2.5 2.5 2

2.5

 Block-based TFT Upload to any peer as long as:

upload download < constant threshold.

FairTorrent and BB-TFT appear similar but usage of deficits is fundamentally different. PropShare In each round peer reallocates upload rates in proportion to observed download rates. Assumes in each round peers can accurately estimate intended rate allocations of all neighbors.

FairTorrent Algorithm
Its

a distributed algorithm that provides fair data exchange, without any global allocation or management service beyond what is already provided by BitTorrent.

A FairTorrent client uploads a data block to the peer it

owes the most data and automatically converges to the individual reciprocation rates of its peers, without measuring or predicting these rates

Leecher Behaviour

Seed Behavior
FairTorrent allocates seed bandwidth to be split evenly

among leechers by simply sending blocks in a round-robin fashion.

Exchanging Data
Unchoking Peers By default, FairTorrent establishes TCPconnections with 50 neighbors. A FairTorrent leecher unchokes each neighbor interested in its data. Reaching Rate Convergence FairTorrent reachs at much faster rate convergence than PropShare

Data Availability
FairTorrent leverages BitTorrents rarest first policy when

requesting a block from a leecher or a seed. Bootstrapping

New peer needs a block to start trading FT seeds send blocks in a round robin fashion to all the

leechers.
Mega-Capacity Nodes
Node whose capacity exceeds the sum of the upload

capacity of its neigbors.

FairTorrent Properties
FairTorrent incurs low overhead. The most expensive step

in a FairTorrent implementation is the reinsertion of a peer inside the SortedPeerList, a step that runs in O(log n). Fairness and Service Error Service error is the max difference betweendifference between the number of bytes a peer has contributed and received from other leechers at any time during a download session.

 Fast Convergence and High Utilization Resilience to Strategic Peers

Generic Strategy WhiteWashing

EXPERIMENTAL RESULTS
FairTorrent (FT) in both the original BitTorrent Python client,

which implements the documented version of the BitTorrent protocol, and in the popular Azureus Java client.
Evaluated on PlanetLab against: Original BitTorrent client Azureus (most popular) PropShare BitTyrant (uses K-TFT with other BitTyrant clients)

 Four sets of experiments: 1) Clients with upload capacities across a uniform distribution, which begin downloading simultaneously and remain as seeds in the system upon download completions. 2) A high-contributing peer participating in a swarm with low contributors and free-riders, all of which begin downloading simultaneously and remain as seeds in the system upon download completions. 3) Clients with upload capacities based on a distribution from live BitTorrent networks, which join the system asynchronously and leave upon download completion. 4) Live BitTorrent swarms.

Test Parameters
32 MB file Default connections

-50 for Azureus -80 for BitTorrent -500 for Propshare and BitTyrant

Uniform Distribution
50 leecher 10 seeds Upload capacity 50kB/s

Download capacity 100kB/s

Five experiments done with five different sets of upload

capacities generated randomly from the respective distribution, resulting in 250 leecher measurements.

Uniform Distribution
FairTorrent BitTorrent

Azureus

BitTyrant

Rate Convergence

Fig :U: High uploaders rate convergence.

Fig:U: Rate convergence for all Peers including free riders.

Max Service Error

Download time

U: High uploaders download time.

U: Download time of all leechers.

Download Time

U: FairTorrent download time.

U: BitTorrent download time.

U: FairTorrent download time.

U: PropShare download time.

Skewed Distribution

S: Download time.

Dynamic Live Distribution

The dynamic live distribution represents peers with upload

capacities taken from live torrents, which asynchronously join and leave, do not seed upon completion if they are leechers, and participate in a larger, nonmesh network.
A larger network of 100 leechers and 20 seeds is used

and configured the upload capacities of the leechers based on a distribution of the upload capacities in live BitTorrent networks

Dynamic Live Distribution

Live swarms

Conclusions
FairTorrent is a deficit-based algorithm avoids the pitfalls

of previous approaches that suffer from slow peer discovery, inaccurate bandwidth estimates, bandwidth underutilization, and complex tuning of parameters.
FairTorrent achieves much more optimal fairness, rate-

convergence and resilience to strategic peers than rateallocation approaches

Guarantees

better performance for high-contributing

peers

REFERENCES
[1] Alex Sherman, Jason Nieh, and Clifford Stein FairTorrent: A Deficit-Based DistriButed Algorithm to Ensure Fairness in Peer-to-Peer ystemsIEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 20, NO. 5, OCTOBER 2012 [2] A. Bharambe, C. Herley, and V. Padmanabhan, Analyzing andimproving a BitTorrent networks performance mechanisms, in Proc.25th IEEE INFOCOM, Apr. 2006, pp. 1 12. [3] B. Fan, D. Chiu, and J. Lui, The delicate tradeoffs in BitTorrent-like file sharing protocol design, in Proc. 14th IEEE ICNP, Nov. 2006, pp.239248. [4] Q. Lian, Y. Peng, M. Yang, Z. Zhang, Y. Dai, and X. Li, Robust incentives via multi-level tit-for-tat, presented at the 5th IPTPS, Feb. 2006