Loading [a11y]/accessibility-menu.js
Analytical Model for BitTorrent-Based Live Video Streaming | IEEE Conference Publication | IEEE Xplore

Analytical Model for BitTorrent-Based Live Video Streaming


First Page of the Article

Date of Conference: 11-13 January 2007
Date Added to IEEE Xplore: 15 May 2007
Print ISBN:1-4244-0667-6

ISSN Information:

Conference Location: Las Vegas, NV, USA

First Page of the Article


I. Introduction

In early peer-to-peer file-sharing systems, files could be exchanged only when peers had the files in their entirety. However, since most users leave the system when the file download completes, a very small fraction of the total peer upload capacity was ever utilized. By fragmenting files into small pieces (and by employing a Tit-for-Tat policy in choosing the peers to which a peer would upload fragments to), BitTorrent [1] allowed utilization of a large fraction of the total peer upload capacity. The idea of exchanging fragments also enabled supporting streaming applications over a peer-to-peer network and several peer-to-peer video streaming applications e.g. CoolStreaming [3] and PPLive [5] have come up in recent years. Measurements on these systems report over 100,000 concurrent users on a single channel for PPLive [6] and over 25,000 concurrent users on a single channel for CoolStreaming [4]. These numbers have generated interest in using this approach for commercial live video streaming. While the scalability of peer-to-peer solutions [8] is very attractive, established providers need to understand the performance of a system utilizing peer-to-peer technology before they deploy it in their systems as users have come to expect a certain level of quality from them and any large deviations from customer expectations would be detrimental to their reputation. In this paper, we provide an analytical model to inform the design of a BitTorrent-based live video streaming solution. Given the asymmetric nature of the broadband access links currently in deployment, it is obvious that a pure peer-to-peer solution can support only limited streaming rates. Our work in this paper allows service providers to estimate the server-side capacity needed to support a targeted number of users at a given streaming rate.

References

References is not available for this document.