Optimal delivery of multi-media content over networks

Abstract

In this paper, we describe scalable optimal methods for delivering archived and live multi-media content from servers to multi-media client players endowed with substantial RAM or disk-based buffers. These methods result from the application of linearoptimization theory (linear programming) to the problem of how best to modulate the flow rate of constant-bit-rate (CBR) content for all sessions linking a server to its clients, in which session flow rates are subject to upper and lower bound constraints, and aggregate flow cannot exceed a specified maximum. An efficient O(n) algorithm to maximize aggregate flow is described. We propose a tunable minimum constraint on session flows that is shown to result in a rapid and sustained accumulation of reserve content within a player's buffer. An associated Call Admission Control (CAC) algorithm is also described. The benefits of the methods described include improved server efficiency, enhanced end-user experience (QOS), cost effective end to-end content delivery, directly from origin servers to clients without need of intervening edge-caching technology.