This paper describes a buffer sharing technique that strikes a balance between the use of disk bandwidth and memory in order to maximize the performance of a video-on-demand server. We make the key observation that the configuration parameters of the system should be independent of the physical characteristics of the data (e.g., popularity of a clip). Instead, the configuration parameters are fixed and our strategy adjusts itself dynamically at run-time to support a pattern of access to the video clips.
/lp/association-for-computing-machinery/buffer-sharing-in-video-on-demand-servers-CeMG9EGix0