Aflevering van gepersonaliseerde multimediale data in peer-to-peer netwerken

Delivery of personalized multimedia content on a peer-to-peer network

Birger Anckaert Supervisor(s): dr. Davy Van Deursen, Erik Mannens, prof. dr. ir. Rik Van de Walle

Abstract – This article will describe the design of a peer­to-peer multimedia streaming protocol optimized for personalized multimedia content distribution. The first phase of the development entails a specification of an optimal peer-to-peer architecture for general multimedia streaming purposes. In a second phase an automated metadata driven framework for multimedia editing is introduced in order to obtain personalized multimedia presentations. The gained multimedia view will then be transported through the streaming infrastructure specified in the first phase of development.

Keywords – peer-to-peer multimedia streaming, personalized multimedia streaming, automated multimedia editing, metadata driven personalization of multimedia.

I. Introduction

Broadband technology advances have allowed the successful deployment of multimedia streaming for video­conferencing and surveillance applications over private networks for quite some time now. However streaming high-quality multimedia over the Internet to address the constantly growing quality demands of theend user still remains a challenging issue. In recent years, peer-to-peer (P2P) technologyhas captured the interest of the research community as well as the industry. By allowing peers to serve each other, P2P solutions overcome many limitations of traditional client-server architectures. Theycan handle flash crowds as well as achieve bandwidth scalability. In addition, P2P solutions do not require any special support from the underlying network topology. As a result, there has been an increasing interestin the use of P2P architecture for large-scale, high-quality multimedia streaming.

With the development of the Semantic Web metadata has become of growing interest to software developers. Many tasks can be automated and integrated through the use of a diligent metadata annotation. By introducing a metadata annotation to multimedia it might be possible to create an automated multimedia editing platform which supports the creation of multimedia presentations solely based on the use of meta information.

When both technologies are combined, a peer-to-peer protocol with streaming capabilities for personalized multimedia content is formed.

The remainder of this paper is structured as follows: Section 2 provides a quick overview of the proposed P2P architecture. Section 3 introduces the inner workings of the personalization mechanism. Finally section 4 has some conclusions about the specification en suggestions for future work.

II. P2P streaming protocol

The proposed P2P streaming architecture is a combination of several strong points of existing architectures. The protocol is closely modeled to the BitTorrent (1) (2) approach, with the main exception of the tracker system. Instead of tracker servers, super users much like the Super Nodes in the FastTrack protocolspecification are used. This way the system has no need for a centralized bootstrappingmechanism. Because of the inherent sequential encoding and playback of multimedia, and the generally larger file sizes of the mediatypes some modifications to the standard P2P transport protocols had to be made in order to accomodate multimedia streaming.

Firstly, due to the parallel transport of multiple multimedia chunks over a besteffort protocol, the order of arrival of the pieces could be disturbed. The introductionof a simple reorder buffer between the arrival of the multimedia fragment and itsplayback should fix this problem. Secondly, in order to secure some progress during the playback of the streaming multimedia a playback buffer is added to anticipate for bandwidth fluctuations. One and the same buffering mechanism can actually be used for both buffer services. The output of the buffer mechanism is fed to a genericmultimedia playback framework in order to view the transferred multimedia presentation.

III. Personalized multimedia

In order to have an idea of the network multimedia content, all shared media already should be annotated with a usable metadata vocabulary. A practical multimedia annotations environment will have to fulfill certain requirements (2). RDF is most able to implement all of these concerns in a suitable multimedia annotation environment, due to its adaptive nature.

Once all annotation needs are tended to, an automated multimedia environment could use a powerful query language (e.g. SPARQL) to isolate multimedia fragments which peak a users’ interest. These fragments can be aggregated in such a way its composition is tailoredspecifically to the needs of one user.

IV. Conclusions

When the specifications proposed in this paper are implemented to the letter, a very rudimentary nevertheless workingprotocol for P2P streaming of personalized multimedia content is gained.

The implementation of the proposed ideas was never meant to be an applicable protocol API. It was intended as a mere proof-of-concept in order to appraise the practical possibility to create such a protocol. The application has never been deployed on more than one machine andwas always tested on a locally emulated network. Suffice to say the application has some shortcomings.

Some future optimizations and additions to the protocol may include butare not limited to: the optimization of the configuration of peer groups, the implementation of an adaptive reorder buffer mechanism, the specification of a practical multimedia annotations environment, and the integration of a full-fledged automated multimedia editing environment.

References

1. Shah, Purvi en Pâris, Jehan-François.

Peer-to-Peer Multimedia Streaming Using BitTorrent. [document] Houston : University of Houston, 2007.

2. Xiaojun, Hei, Lui, Yong en Ross, Keith

W. IPTV over P2P Streaming Networks: the Mesh-pull Approach. [document] Brooklyn : Polytechnic University, 2008.

3. Geurts, Joost, van Ossenbruggen, Jacco en Hardman, Lynda. Requirements for practical multimedia annotation.

[document] Amsterdam : CWI, 2005.