Practical real-time approaches for scheduling VBR video in a client-server computing environment

Scheduling of multiple variable-bit-rate (VBR) video packet streams at the server is one of the most crucial issues in the design of a enterprise video-on-demand (VoD) system. Scheduling is important to guarantee the requested QoS levels while maximizing the utilization of the resources. But the scheduling problem is difficult due to the inherent varying resource requirements of the video streams. Although much work on smoothing and scheduling algorithms has been done, these previous techniques, which usually require the knowledge of future frame sizes and are of a high time complexity, are not suitable because in the video application environment envisioned in our study, real-time response from the server is required. In this paper, we propose and evaluate four real-time approaches using real MPEG data. Our results reveal that the proposed real-time techniques introduce only a slight jitter effect on the video display. Furthermore, time-division-based algorithms are superior to rate-based ones.