| Paper Title |
A Novel Cache-aided Fog-RAN Architecture |
| Paper Identifier | FR4.R1.1 |
|---|
| Authors |
Kai Wan, Technische Universität Berlin, Germany; Daniela Tuninetti, University of Illinois at Chicago, United States; Mingyue Ji, University of Utah, United States; Giuseppe Caire, Technische Universität Berlin, Germany |
| Session |
Coded Caching V |
| Location |
Le Théatre (Parterre), Level -1 |
| Session Time |
Friday, 12 July, 16:40 - 18:00 |
| Presentation Time |
Friday, 12 July, 16:40 - 17:00 |
| Manuscript |
Click here to download the manuscript |
| Abstract
| This paper considers a novel cache-aided Fog Radio Access Network (Fog-RAN) architecture including a Macro-cell Base Station (MBS), several Small-cell Base Stations (SBSs), and users. Some users, not in the reach of any SBS, are directly served by the MBS, while the other users are ``offloaded'' and receive information only from the SBSs through high throughput links. In order to alleviate the load in the wireless front-haul links between the MBS and the SBSs, caching is employed at the SBSs. The MBS sends coded packets to the SBSs and to the directly served users via wireless multicast transmission on a common downlink channel, modeled as an error-free shared link of fixed capacity. Subsequently, the SBSs communicate among one another in a Device-to-Device (D2D) fashion so as each SBS obtains enough information to decode the files demanded by its connected users. The access links between SBSs and users are assumed to operate at a sufficiently high rate such that they are not the system bottleneck. For this novel Fog-RAN model, the memory-loads tradeoff for the worst-case demands is investigated. The main contributions of this paper are: (i) a novel symmetric inter-file coded cache placement scheme, (ii) a novel D2D delivery scheme to handle the inter-SBS communication phase, that is order optimal when each SBS serves the same number of users, and (iii) a novel asymmetric cache placement with file subpacketization dependent on the network structure, which is exactly optimal in some memory size regimes. |
