Abstract

Multimedia applications are often considered adaptive in the sense that their behaviour can be adapted to the available resources. One way to achieve this is through a selective dropping of media frames. Different media types have different tolerance to loss, for example, humans are more sensitive to loss of continuity in audio as compared to video. It is, therefore, attractive to develop best effort services targeted for a particular media type (e.g., video). Such a service should provide graceful performance degradation as resources become scarce by exploiting the particular characteristics of the media type. In this thesis, we consider providing a best effort service targeted for video transmission, by multiplexing multiple video streams over a reserved bandwidth network channel. The emerging services architecture in ATM networks as well as the Internet provide the capability to reserve resources such as bandwidth, and to use them to provide higher level services. We use simulation to consider the effect of different scheduling and buffer management policies on the performance when the bandwidth is scarce and losses are inevitable. By using effective frame rate as a performance measure, we show that FIFO scheduling is unable to ensure fairness in terms of equitable performance degradation of the multiplexed streams. Our simulations reveal that explicit bandwidth allocation is required, and that this bandwidth allocation must be (1) targeted to ensure that the performance of each of the multiplexed streams degrades equitably, and (2) must be dynamically changed to take advantage of the time varying characteristics of video data