Bulk prefetching with deadline-driven scheduling to minimize startup latency of continuous media servers

Researches on the continuous media (CM) servers have focused on the throughput issues because the most popular CM application has been the video-on-demand system which provides relatively long video clips or movies. Latency issue has received little atten-tion in such a system because a few seconds up to a few minutes latency is considered tolerable. However, as a variety of multime-dia applications are emerging, some applications begin to require interactive presentation of continuous media where the most chal-lenging issue is to minimize startup latency, i.e., minimizing the time between the issuing a request and the start of the display. This paper proposes bulk prefetching techniques to minimize startup latency and analyzes their performance on multi-disk CM servers utilizing a deadline-driven scheduling and random data placement. Simulation results demonstrate a fraction of a second startup la-tency, which is acceptable for most interactive applications. 1