In this paper, we analyze the relative merits of uniform versus priority dropping for the transmission of layered video. We first present our original intuitions about these two approaches, and then investigate the issue more thoroughly through simulations and analysis in which we explicitly model the performance of layered video applications. We compare both their performance characteristics and incentive properties, and find that the performance benefit of priority dropping is smaller than we expected, while uniform dropping has worse incentive properties than we previously believed.

A key question in the definition of new services for the Internet is whether to provide a single class of relaxed real-time service or multiple levels differentiated by their delay characteristics. In that context we pose the question: is service priority useful in networks? We argue that, contrary to some of our earlier work, to properly address this question one cannot just consider raw network-centric performance numbers, such as the delay distribution. Rather, one must incorporate two new elements into the analysis: the utility functions of the applications (how application performance depends on network service), and the adaptive nature of applications (how applications react to changing network service). This last point is especially crucial; modern Internet applications are designed to tolerate a wide range of network service quality, and they do so by adapting to the current network conditions. Most previous investigations of network performance have neglected to include this adaptive behavior.In this paper we present an analysis of service priority in the context of audio applications embodying these two elements: utility functions and adaptation. Our investigation is far from conclusive. The definitive answer to the question depends on many factors that are outside the scope of this paper and are, at present, unknowable, such as the burstiness of future Internet traffic and the relative offered loads of best-effort and real-time applications. Despite these shortcomings, our analysis illustrates this new approach to evaluating network design decisions, and sheds some light on the properties of adaptive applications.