On the Average Performance of Caching and Coded Multicasting with Random Demands

TL;DR

This paper introduces a new decentralized caching and coded multicasting scheme optimized for random i.i.d. user demands following a Zipf distribution, achieving the first order-optimal performance in this setting.

Contribution

It proposes a novel decentralized random caching and coding scheme that is proven to be order-optimal for average demand scenarios with Zipf distribution.

Findings

Achieves order-optimal average transmission performance.

First to demonstrate order-optimality for random demand caching.

Effective for Zipf-distributed user demands.

Abstract

For a network with one sender, $n$ receivers (users) and $m$ possible messages (files), caching side information at the users allows to satisfy arbitrary simultaneous demands by sending a common (multicast) coded message. In the worst-case demand setting, explicit deterministic and random caching strategies and explicit linear coding schemes have been shown to be order optimal. In this work, we consider the same scenario where the user demands are random i.i.d., according to a Zipf popularity distribution. In this case, we pose the problem in terms of the minimum average number of equivalent message transmissions. We present a novel decentralized random caching placement and a coded delivery scheme which are shown to achieve order-optimal performance. As a matter of fact, this is the first order-optimal result for the caching and coded multicasting problem in the case of random demands.