Current Web based systems are designed for the static class of document based information. In the next generation of information systems, digital video will form an important class of traffic. Emerging systems will need to handle multimedia applications that possess varying traffic characteristics and have Quality-of-Service(QoS) requirements in terms of bounded delay, jitter, loss rate, synchronization skew etc. [14]. In order to design and deploy these complex and evolving systems with proper cost/performance ratios, we must understand their expected behaviors and workloads. Large scale video service applications that are currently being deployed have some significant problems: (1) User dissatisfaction due to poor QoS. (2) Poor cost-performance ratios due to inefficient management of system resources, especially when guaranteed service is desired. In order to resolve these issues, we will initially need to identify bottlenecks in the system that are responsible for poor response times. We can then determine suitable mechanisms to obtain cost-effective QoS.
In this paper, we address the issues in designing metrics that are important in evaluating the QoS of video transmission. There has been little work in determining effective metrics of QoS for video transmission that characterize both cost (revenue generated or service demand) and guaranteed service. The metrics of analysis and comparison for video transmission must be determined as an end-to-end measure of QoS from video server to end-user(s). By developing these metrics, we hope to enhance the client, server and networking components of a system with monitoring capabilities to measure and evaluate video characterizations. This paper is organized as follows. In Section 2, we discuss a workload model for developing and understanding QoS metrics. Section 3 presents empirical studies and experimental justification for the metric selection based on the three systems - VOSAIC, hierarchical VOD and the remote VCR systems. Section 4 proposes a new integrated metric for measuring video QoS and the analytical framework to express the tradeoffs. We also propose a metric-based QoS architecture along with negotiation and reward protocols. In Section 5 we discuss related work and conclude with future research directions in Section 6.