TEST ENVIRONMENT FOR PERFORMANCE EVALUATION OF AN INTERNET RADIO David Melendi, Manuel Vilas, Xabiel G. Pañeda, Roberto García, Víctor G. García Computer Science Department, University of Oviedo Campus Universitario de Viesques. Sede Departamental Oeste, 33204 Xixón-Gijón, Asturies Keywords: Audio, Evaluation, Internet, Live, Multimedia, Radio, Streaming. Abstract: This paper presents a test environment designed to improve Internet radio services through the evaluation of different service features. The environment comprises the generation of audio streams, the delivery of those streams through different communication networks, and the access of final users to those contents. A broad set of service architectures can be emulated, and several network configurations can be deployed using the available communication devices. It is also possible to simulate users’ behaviour thanks to a workload generator that can be configured using statistical information obtained from real service access data. A case study is also presented, where a glimpse of the possibilities of the environment can be caught. This test environment will allow service administrators or research teams to predict what will happen in a real service if its configuration is modified, or if user behaviour changes. Furthermore, managers will be capable of knowing whether an Internet radio service can be improved or not. 1 INTRODUCTION The emergence of the World Wide Web has changed the Internet world. This service has become a powerful medium. Daily, millions of users in the world browse the web, produce an important number of accesses, and a huge volume of information is delivered to them. Like other communication enterprises, traditional Radio companies have discovered the Internet to be an important tool to reach millions of users in the world without any constraint related to commercial licenses, limited frequency spectrums, high infrastructure costs, etc. Now technology is not an issue for these companies. The bandwidth increase in subscribers’ access capabilities, and the development of the streaming technology have given rise to the appearance of a new complementary service: the Internet Radio. There are two types of audio services on the Internet: live-audio and jukebox (audio-on-demand). In jukebox services, the user requests the information at any time and the server delivers it exclusively. This system allows users to interact with information: Pauses, backward and forward jumps are allowed. Its behaviour is similar to a CD player, and it is the user who controls what to listen to in every moment: these are user driven services (Veloso, 2002). On the other hand, live-audio is more similar to a traditional radio service. Contents are received directly by a multimedia server, which broadcasts them straight out to the audience. In these services users cannot choose what to listen to, as there is a previously established programme. These are object driven services (Veloso, 2002). Most radio services on the Internet are based on streaming technology. The advantages of audio streaming and the subscribers’ expectations are important. However, this technology presents some problems. Although audio delivering is not so resource-consuming as video, it also requires a constant quality of service. What is more, live-audio services require greater transmission capabilities than jukebox services, due to the fact that they tend to suffer stress periods depending on programme evolution, known as prime-time periods. In these stress periods a wide range of users connect at the same time to the same source of contents. To maintain service quality under control, it is important to select the best configuration parameters possible. These parameters include, among others, the quality of the produced contents, the architecture of the service, the configuration of the network, etc. Nowadays, all these parameters are usually established by service managers based on their own experience. These managers usually know what the most appropriate configuration is, what qualities must be generated, or where to place each service device. Nevertheless, each service is usually 183 Melendi D., Vilas M., Pañeda X., Garcia R. and Garcia V. (2005). TEST ENVIRONMENT FOR PERFORMANCE EVALUATION OF AN INTERNET RADIO. In Proceedings of the Second International Conference on e-Business and Telecommunication Networks, pages 183-190 Copyright c  SciTePress