Bit Torrnt
Bit Torrnt
Bit Torrnt
Proportion of leechers
Proportion of seeders
0.7 0.3
0.6 0.6 0.4
0.5 0.5
0.4 0.4 0.6
0.3 0.7
0.2 0.2 0.8
0.1 0.9
0 0 1
0 500 1000 1500 2000 2500 3000 2000 3000 4000 5000 6000 7000
Number of connecting clients Time (s)
Figure 2: Evolution of the simulation time as a func- Figure 4: Simulation data representing the propor-
tion of the amount of connecting nodes during a sim- tion of seeders and leechers in the swarm during a
ulation run. simulation run.
In Figure 3, we show the Empirical Probability Density rate is set in the initialisation file. The session duration time,
Function (EPDF) for the piece distribution for a simulation however, is partially dependant on the amount of peers in
scenario with 1000 peers. The grey line is the theoretical the swarm and the initial amount of pieces of the peers as
density for a uniform distribution with minimum value 0 well as the piece response delay and available bandwidth. In
and maximum value 1204, while the black line represents order to be able to compare similar data of simulation re-
the proportion of requests of the associated piece. sults and real-world measurements we achieved a simulated
swarm size similar to the one observed in the real-world
measurements [9].
Piece distribution.
Figure 5(a) shows a graph of the session duration Com-
1e−03
5. FUTURE WORK
0 200 400 600 800 1000 1200 While our simulator is capable of handling a wide range
Piece number of scenarios, there is still some functionality missing. For
instance, modular peer selection is not implemented, nor
is peer snubbing, i. e. , dropping peers that do not respond
Figure 3: Piece distribution for 1000 peer scenario. quickly enough. Also, as the BT protocol allows for exten-
sions, several clients have used this capability to add new
As Figure 3 shows, the simulated piece selection algorithm features to the protocol, e. g. , a trackerless Distributed Hash
quite clearly tends toward a uniform selection of pieces. Table (DHT) protocol, encryption, super seeding. Addition-
ally, some clients have extended the torrent file format to
4.5.2 Choking algorithm
contain more information, such as current download status,
The proper working of the choking algorithm was checked connected peers and QoS information. None of these non-
by tracing what decisions each peer made while running the standard extensions have been implemented in our simula-
choking algorithm. No irregularities were observed during tor. In the near future, we plan to add both snubbing and
our simulation test-runs. modular peer selection capabilities to our simulator, as well
as extend the protocol with QoS-related messaging.
4.5.3 Number of peers in swarm Furthermore, points of improvement can be carried out to
Figure 4 shows the amount of seeders and leechers in the improve the performance of the BT simulator. The current
swarm, taken from a simulation run. There are no noticeable version of the simulator loads all the BT clients that will
errors observed in this plot, and the results shows similarities connect during the simulation at the simulation start-up.
with the corresponding results in [11]. We will change this to loading each client when the session
4.5.4 Session duration arrives in simulation time. This way we will be able to simu-
Realistic session arrivals and session duration time are late larger and longer simulations. Further, an overloadable
fundamental for a realistic simulation. The session arrival piece selection algorithm would be preferred instead of a sim-
ple hard-coded switch. Also, the TorrentFile module needs
improvement so it can load more .torrent-files.
Additionally, we will run more and longer simulation runs,
as well as repeat runs with varying initial random number
seeds to achieve confidence intervals for the simulation re-
sults.
6. ACKNOWLEDGMENTS
We would like to thank the Swedish Internet Infrastruc-
ture Foundation (IIS) and Euro-NGI for granting and sup-
porting the ROVER project during 2006 and 2007.
7. REFERENCES
[1] I. Baumgart, B. Heep, and S. Krause. OverSim: A
0
80.0%
90.0%
95.0%
99.0%
[3] A. R. Bharambe, C. Herley, and V. N. Padmanabhan.
Analyzing and improving a bittorrent network’s
3
performance mechanisms. In INFOCOM 2006.
log x
Proceedings of the 25th IEEE International Conference
(a) Simulated results. on Computer Communications., pages 1–12, 2006.
[4] BitTorrent, Inc. Bittorrent. http:www.bittorrent.com.
URL verified on April 10, 2007.
[5] B. Cohen. BitTorrent protocol specification.
0
http://www.bitconjurer.org/BitTorrent/protocol.html,
February 2005.
[6] B. Cohen. BitTorrent. http://www.bittorrent.com/,
−1
March 2006.
log P[X ≥ x]
95.0%
99.0% video-on-demand. In Proceedings of IEEE 7th
1 2 3 4 5 Workshop on Multimedia Signal Processing, pages 1–4,
log x
October 2005.
[9] D. Erman. Bittorrent traffic measurements and
(b) Measured result. models, October 2005. Licentiate thesis, Blekinge
Institute of Technology.
Figure 5: BT session duration. [10] D. Erman. Extending bittorrent for streaming
applications. In Proceedings of the 4th Euro-FGI
workshop on ”New Trends in Modelling, Quantitative
Methods and Measurements”, May/June 2007.
[11] M. Izal, G. Urvoy-Keller, E. Biersack, P. Felber, A. A.
Hamra, and L. Garcés-Erice. Dissecting BitTorrent:
Five months in a torrent’s lifetime. In Passive and
Active Measurements (PAM2004), 2004.
[12] M. Jelasity, A. Montresor, G. P. Jesi, and S. Voulgaris.
Peersim. http://peersim.sourceforge.net, October
2007.
[13] P. Korathota. Investigation of swarming content
delivery systems. Master’s thesis, Sydney University of
Technology, http://me55enger.net/swarm/thesis.pdf,
November 2003.
[14] A. Legout, G. Urvoy-Keller, and P. Michiardi. Rarest [18] A. Sharma, A. Bestavros, and I. Matta. dPAM: A
first and choke algorithms are enough. In ACM Distributed Prefetching Protocol for Scalable
SIGCOMM/USENIX ICM’2006, October 2006. Asynchronous Multicast in P2P Systems. In
[15] S. Naicken, B. Livingston, A. Basu, S. Rodhetbhai, Proceedings of Infocom’05: The IEEE International
I. Wakeman, and D. Chalmers. The state of Conference on Computer Communication, Miami,
peer-to-peer simulators and simulations. SIGCOMM Florida, March 2005.
Comput. Commun. Rev., 37(2):95–98, 2007. [19] The ESM Project. ESM – end system multicast.
[16] PeerCast.org. Peercast – p2p casting for everyone. Online at http://esm.cs.cmu.edu/. URL verified on
Online at http://peercast.org. URL verified on March March 26, 2007.
27, 2007. [20] A. Vargas. OMNeT++ discrete event simulation
[17] D. Qiu and R. Srikant. Modeling and performance system. http://www.omnetpp.org.
analysis of bittorrent-like peer-to-peer networks. [21] A. Vlavianos, M. Iliofotou, and M. Faloutsos. BiToS:
Technical report, University of Illinois at Enhancing BitTorrent for supporting streaming
Urbana-Champaign, USA, 2004. applications. In 9th IEEE Global Internet Symposium
(GI2006), Barcelona, Spain, April 2006.
[22] W. Yang and N. Abu-Ghazaleh. Gps: A general
peer-to-peer simulator and its use for modeling
bittorrent. mascots, 00:425–434, 2005.