An Efficient Coded Multicasting Scheme Preserving the Multiplicative Caching Gain

TL;DR

This paper analyzes the limitations of existing coded multicasting schemes in finite packet regimes and proposes a new algorithm that recovers much of the caching gain without requiring asymptotic conditions.

Contribution

It extends the asymptotic analysis to heterogeneous settings and introduces a polynomial-time graph coloring algorithm to improve practical caching performance.

Findings

Existing schemes lose caching gain at finite packetization

The new algorithm recovers significant caching gain

Order-optimal performance is achievable in practical regimes

Abstract

Coded multicasting has been shown to be a promis- ing approach to significantly improve the caching performance of content delivery networks with multiple caches downstream of a common multicast link. However, achievable schemes proposed to date have been shown to achieve the proved order-optimal performance only in the asymptotic regime in which the number of packets per requested item goes to infinity. In this paper, we first extend the asymptotic analysis of the achievable scheme in [1], [2] to the case of heterogeneous cache sizes and demand distributions, providing the best known upper bound on the fundamental limiting performance when the number of packets goes to infinity. We then show that the scheme achieving this upper bound quickly loses its multiplicative caching gain for finite content packetization. To overcome this limitation, we design a novel polynomial-time algorithm based on random greedy graph- coloring that, while keeping the same finite content packetization, recovers a significant part of the multiplicative caching gain. Our results show that the order-optimal coded multicasting schemes proposed to date, while useful in quantifying the fundamental limiting performance, must be properly designed for practical regimes of finite packetization.