Performance analysis of data placement and retrieval schemes for video servers based on response time guaranteed

A crucial QoS parameter, initial response time for a new request, was often neglected while designing or evaluating the storage subsystem of a video server in previous studies. For different types of services such as interactive video-on-demand, near video-on-demand, and pay-per-view systems, the initial delay of a new request can vary from 0.5 seconds to 5 minutes. This issue has a deep impact on choosing an efficient disk scheduling and data placement scheme for a particular type of video service. In this paper, a new performance model is proposed to evaluate the efficiency of a disk-based video server, in which both factors, number of supported streams and allowable response time, for new requests are considered simultaneously. According to the proposed model, the most efficient storage configuration can be easily obtained under the given parameters.