Cache-aided combination networks with interference

Abstract

Centralized coded caching and delivery isstudied for a radio access combination network (RACN),whereby a set ofHedge nodes (ENs), connected to acloud server via orthogonal fronthaul links with limitedcapacity, serve a total ofKuser equipments (UEs) overwireless links. The cloud server is assumed to hold alibrary ofNfiles, each of sizeFbits; and each user,equipped with a cache of sizeμRNFbits, is connectedto a distinct set ofrENs each of which equipped witha cache of sizeμTNFbits, whereμT,μR∈[0,1]arethe fractional cache capacities of the UEs and the ENs,respectively. The objective is to minimize the normalizeddelivery time (NDT), which refers to the worst case deliverylatency when each user requests a single distinct file fromthe library. Three coded caching and transmission schemesare considered, namely theMDS-IA,soft-transferandzero-forcing (ZF)schemes. MDS-IA utilizes maximum distanceseparable (MDS) codes in the placement phase and realinterference alignment (IA) in the delivery phase. Theachievable NDT for this scheme is presented forr= 2and arbitrary fractional cache sizesμTandμR, and alsofor arbitrary value ofrand fractional cache sizeμTwhen the cache capacity of the UE is above a certainthreshold. The soft-transfer scheme utilizes soft-transferof coded symbols to ENs that implement ZF over the edgelinks. The achievable NDT for this scheme is presentedfor arbitraryrand arbitrary fractional cache sizesμTandμR. The last scheme utilizes ZF between the ENs andthe UEs without the participation of the cloud server inthe delivery phase. The achievable NDT for this scheme is presented for an arbitrary value ofrwhen the totalcache size at a pair of UE and EN is sufficient to store thewhole library, i.e.,μT+μR≥1. The results indicate thatthe fronthaul capacity determines which scheme achievesa better performance in terms of the NDT, and thesoft-transfer scheme becomes favorable as the fronthaulcapacity increases.