2017/01/24 by Kai Wan, Wan, Kai, Daniela Tuninetti +5
Computer Science · Engineering · #Advanced Wireless Communication Technologies #Caching and Content Delivery #Cooperative Communication and Network Coding #FOS: Computer and information sciences #Information Theory (cs.IT)
paper · pdf · doi:10.48550/arxiv.1701.06884
openalex publication_date 2017/01/24 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Caching is an efficient way to reduce network traffic congestion during peak\nhours by storing some content at the users' local caches. For the shared-link\nnetwork with end-user-caches, Maddah-Ali and Niesen proposed a two-phase coded\ncaching strategy. In practice, users may communicate with the server through\nintermediate relays. This paper studies the tradeoff between the memory size\nM and the network load R for networks where a server with N files is\nconnected to H relays (without caches), which in turn are connected to K\nusers equipped with caches of M files. When each user is connected to a\ndifferent subset of r relays, i.e., K = binomHr, the system is\nreferred to as a it combination network with end-user-caches.\n In this work, converse bounds are derived for the practically motivated case\nof it uncoded cache contents, that is, bits of the various files are\ndirectly pushed into the user caches without any coding. In this case, once the\ncache contents and the user demands are known, the problem reduces to a general\nindex coding problem.This paper shows that relying on a well-known "acyclic\nindex coding converse bound" results in converse bounds that are not tight for\ncombination networks with end-user-caches. A novel converse bound that\nleverages the network topology is proposed, which is the tightest converse\nbound known to date. As a result of independent interest, an inequality that\ngeneralizes the well-known sub-modularity of entropy is derived. Several novel\ncaching schemes are proposed, based on the Maddah-Ali and Niesen cache\nplacement. The proposed schemes are proved: (i) to be (order) optimal for some\n(N,M,H,r) parameters regimes under the constraint of uncoded cache placement,\nand (ii) to outperform the state-of-the-art schemes in numerical evaluations.\n