An efficient flow control plan for end-to-end delivery of pre-stored compressed videos

This paper studies the end-to-end flow control strategy of delivering videos over a data path, where on one end, a server retrieves large fixed-sized video blocks from its disk drive at a constant block rate. on the other end, a client consumes variable-sized compressed video frames at a fixed frame rate. Between the two ends, a network channel transmits small fixed-sized cells at a constant cell rate. We propose a flow control plan with two hierarchical phases, namely, the peer-level and network-level phases. Our control plans consume both the minimum buffer space and the minimum network channel bandwidth. We conduct a set of experiments by tracing several typical video tracks, and observe a few common facts that for each video flow, (1) the client only need to enforce couple of seconds start-up delay before playback (2) the minimum total buffer space required is less than 1.4 Mbytes and (3) the minimum network bandwidth required is very close to the video´s average bit rate