Fundamental Limits of Caching in Wireless D2D Networks

Abstract

We consider a wireless Device-to-Device (D2D) network where communication is restricted to be single-hop. Users make arbitrary requests from a finite library of files and have pre-cached information on their devices, subject to a per-node storage capacity constraint. A similar problem has already been considered in an "infrastructure" setting, where all users receive a common multicast (coded) message from a single omniscient server (e.g., a base station having all the files in the library) through a shared bottleneck link. In this work, we consider a D2D "infrastructure-less" version of the problem. We propose a caching strategy based on deterministic assignment of subpackets of the library files, and a coded delivery strategy where the users send linearly coded messages to each other in order to collectively satisfy their demands. We also consider a random caching strategy, which is more suitable to a fully decentralized implementation. Under certain conditions, both approaches can achieve the information theoretic outer bound within a constant multiplicative factor.In our previous work, we showed that a caching D2D wireless network with one-hop communication, random caching, and uncoded delivery (direct file transmissions), achieves the same throughput scaling law of the infrastructure-based coded multicasting scheme, in the regime of large number of users and files in the library. This shows that the spatial reuse gain of the D2D network is order-equivalent to the coded multicasting gain of single base station transmission. It is therefore natural to ask whether these two gains are cumulative, i.e., if a D2D network with both local communication (spatial reuse) and coded multicasting can provide an improved scaling law. Somewhat counterintuitively, we show that these gains do not cumulate (in terms of throughput scaling law). This fact can be explained by noticing that the coded delivery scheme creates messages that are useful to multiple nodes, such that it benefits from broadcasting to as many nodes as possible, while spatial reuse capitalizes on the fact that the communication is local, such that the same time slot can be re-used in space across the network.Unfortunately, these two issues are in contrast with each other.

The Throughput-Outage Tradeoff of Wireless One-Hop Caching Networks

“…While in the D2D network transmissions should be "as local as possible" in order to exploit spatial reuse, in the coded multicast network transmissions should be as global as possible, in order to benefit the largest number of users. In a recent follow-up paper [40], we have investigated a decentralized version of network-coded scheme of [16] for the same D2D network of the present work, without any omniscient node that has the whole file library. It turns out that coded multicasting gain and spatial reuse gain do not cumulate.…”

“…Again, following a perturbation argument similar to Appendix J), for (40). Otherwise, the problem is infeasible.…”

The exact rate-memory tradeoff for caching with uncoded prefetching

“…Several coded caching schemes have been proposed since then [11]- [16]. The caching problem has also been extended in various directions, including decentralized caching [17], online caching [18], caching with nonuniform demands [19]- [22], hierarchical caching [23]- [25], device-to-device caching [26], cache-aided interference channels [27]- [30], caching on file selection networks [31]- [33], caching on broadcast channels [34]- [37], and caching for channels with delayed feedback with channel state information [38], [39]. The same idea is also useful in the context of distributed computing, in order to take advantage of extra computation to reduce the communication load [40]- [44].…”

Cache-Aided Content Delivery Over Erasure Broadcast Channels

“…Several improved coded caching schemes and information theoretic performance bounds have been introduced since then [8]- [15]. Coded caching has since been extended to various network settings, including device-to-device caching [16], [17], online cache placement [18], files with non-uniform popularities [19], [20] and distinct lengths [21], users with non-uniform cache sizes [22], [23], multi-layer caching [24], and caching by users with different distortion requirements [25]- [27].…”