On the Locality of BitTorrent-based

Video File Swarming

Haiyang Wang

.

Jiangchuan Liu

.

Ke Xu

.

.

{hwa17, jcliu}@cs.sfu.ca

School of Computing Science, Simon Fraser University, British Columbia, Canada

.

xuke@csnet1.cs.tsinghua.edu.cn

Department of Computer Science and Technology, Tsinghua University, Beijing, China

Abstract¡ª In the past few years, there have been tremendous

interest in the peer-to-peer(P2P) content delivery. Although this

communication paradigm does not need a dedicated server infras-

tructure, it dramatically increases the traffic over inter-ISP links.

In particular, the most popular P2P application, BitTorrent(BT)

generates a huge amount of traffic on the Internet.

To address this challenge, P2P locality has been examined,

which explores the access to local resources to optimize the inter-

ISP traffic. However, most of these approaches have focused on

a global strategy, and attempted to change the peer selection

mechanism, which potentially affects the random topology of

BT and thus reduces its robustness. The content and the

peer diversities are seldom discussed, particularly the video file

swarms of distinct characteristics.

In this paper, we for the first time examine the different BT

contents and peer properties in regards to the locality issues

through a large-scale measurement. We demonstrate the distinct

characteristics of video file swarms, and find that the distribution

of the AS clusters (a set of peers belonging to the same AS)

follows the Mandelbrot-zipf law. Our results also suggest that

the peer in a few ASes are more likely to form large AS clusters

and most ASes on the Internet do not have enough potential for

locality. Therefore, a global locality approach may not be our

best choice. We then address the problem through a selective

locality approach based on a novel peer prediction method.

I . I

N TROD U CTI ON

Peer-to-peer content delivery has become one of the most

popular applications in recent years. BitTorrent, the most

successful P2P file sharing system over the Internet, has been

widely used for the distribution of large files. Although the

P2P paradigm does not have to maintain a dedicated server

infrastructure, it generates a huge amount of traffic over inter-

ISP links. In particular, even though some BT peers are located

in the same or nearby ISPs and downloading identical contents,

they are unnecessarily connected through remote peers. Since

the ISPs typically pay their peering or higher-level ISPs for

global connectivity, the traffic between different ISPs is costly

and presents significant network engineering challenges. To

make the matter worse, the success of BitTorrent has also

greatly motivated the design of new traffic-intensive applica-

tions, such as streaming service, over the Internet [1]. In fact,

BitTorrent itself has already been extensively used for video

file distribution, albeit in a download-and-play mode.

To alleviate the inter-ISP traffic problem, many solutions

have been proposed beyond the straightforward blocking of the

P2P traffic [2]. Among them, P2P locality [3] has been widely

suggested, which explores access localities to reduce long-haul

traffic. There is no doubt about the existence and benefit of

BT locality and there have been a series of implementations.

However, most of them are based on the modification of the

global peer selection mechanism; the peer and content features

are largely ignored during the locality process. Although the

peers may gain some benefits from this modification, the

robustness of the BT networks may be sacrificed due to the

change of random topology. On the other hand, the distinct

characteristics of different contents, particularly video file

swarms, have yet to be measured and explored.

In this paper, we for the first time examine the BT locality

problem with content and the peer diversities through a large-

scale measurement. Our study suggests that the video contents

are obviously quite popular in BT networks. Most of these

video file swarms contain very large files that pose significant

challenges to ISPs. On the other hand, based on the AS level

measurements, we find an interesting relationship between

different ASes by holding the peers that willing to download

the identical contents. Our investigation indicates that the peers

belonging to some ASes are more likely to appear in a large

AS cluster. We further observe a Mandelbrot-zipf distribution

[4] in the ratio of AS cluster size to swarm size (the total

number of peers in a swarm), which indicates that the peers

belonging to a few large AS clusters are indeed more eligible

to be adjusted by a locality mechanism. Therefore, a selective

locality mechanism is required to optimize the overhead and

the robustness of the BT networks.

The rest of this paper is organized as follows: In Section

II, we illustrate the related works. Section III presents our AS

level measurement results. After the description of the main

challenge, we discuss an AS relationship based approach in

section IV to predict whether a peer is likely to appear in a

large AS cluster. Finally, the paper is concluded in Section V.

II. B

ACK G RO U ND A N D

R

EL AT ED

W

OR K

P2P locality attracted attention from many researchers in

recent years. The pioneer work of T. Karagiannis et al. [3] is

the first study to address the locality issues in P2P systems.

Aiming to solve the inter-ISP traffic problem, they studied

both the real traces and simulation results. They also evaluated

1