Video server retrieval scheduling for variable bit rate scalable video

Abstract

In advanced multimedia networks, video servers with real time operating systems and high performance storage architectures will deliver multiple concurrent video streams to clients. The video server has a limited disk I/O bandwidth. This research presents a new framework for optimally utilizing the on board memory to alleviate the disk bandwidth bottleneck. The optimal utilization of the memory resource is achieved by using an optimal retrieval schedule for video data across the disk memory interface. The optimal retrieval schedule is directly applicable to video servers that are based on general storage architectures (e.g. RAID). Performance evaluation using real MPEG2 trace data verifies that this approach can provide large increases in the number of supported video streams.