Coded Caching for Heterogeneous Wireless Networks with Multi-level Access.
Jad Hachem, Nikhil Karamchandani, Suhas N. Diggavi · arXiv (Cornell University) · 2014
Content distribution networks, enabled through replicated caching, have enabled broadband content delivery over the wired Internet. In this paper we propose a wireless content distribution scheme based on a new multi-level storage, access, and distribution architecture that is tailored for next generation wireless networks. This scheme significantly outperforms conventional translation of ideas from the wired Internet to wireless. Our starting point is the observation that wireless network architecture is converging towards a dense deployment of wireless access points (APs) with small coverage, combined with cellular base-stations (BS) with larger coverage. The consequence of this heterogeneous architecture is that a user could access the BS as well as multiple APs. Therefore an optimal design is to trade-off the cost (transmission rate) at the BS, with the caching cost (memory) at the APs and the access cost of connecting to multiple APs at the user, for a given content popularity profile. In this paper we propose a caching scheme with provable performance trading off these costs and theoretically demonstrate its order optimality with respect to information-theoretic bounds. Our solution is to use coded cache storage and the inherent broadcast property of wireless through network coding. We also validate the scheme on several real datasets and quantify the trade-off in rate, memory, and access. These numerical evaluations also suggest an intriguing possibility that categorizing content into a small number of popularity levels is sufficient. I.