A QoS Based Scheduling Algorithm for Continuous Media Servers

A growing set of applications require digital audio and video data, commonly referred to as continuous media (CM). These applications require continuous real{time data ow during play-back and recording. Scheduling strategies to service accesses by multiple concurrent applications at CM data servers (CMS) must provide timely access to storage devices. Besides real{time access, such data is characterized by its large size and high transmission bandwidth and must be compressed to elimiate temporal and spatial redundancy. However, by compressing data, data rate variability is introduced which creates problems to the scheduling strategy like low resource and bandwidth utilization. In this paper we present a scheduling strategy at the CMS for compressed streams called QBSCAN which exploits clients' concessions in playback quality to achieve higher utilization of the CMS. By stating the needed Quality of Service (QoS), clients declare their level of acceptance of tardiness and smoothness during playback. QBSCAN uses clients' QoS speciications to allocate I/O bandwidth and consequently increases the resource utilization. Simultaneously, the total capacity (the number of concurrent streams supported) of the CMS is enhanced. The eecacy of QBSCAN is demonstrated via simulation studies.