Resource Cost Balancing with Caching in C-RAN

Abstract

Cloud radio access networks (C-RAN) are expected to be the backbone of next generation communication networks. In order to reduce the backhaul cost, which is a bottleneck in C-RAN, popular files are cached in local memories at remote radio heads (RRH's) and thus in close proximity to the users demanding it. In this paper we investigate different caching strategies with the objective to reduce backhaul and transmit power cost. It turns out that the problem of jointly minimizing the transmit power and backhaul costs constitutes a mixed integer non linear program (MINLP). First, we introduce slack variables to formulate the problem as a standard mixed integer second order cone program (MI-SOCP). With this formulation we can get the global optimal solution with reduced computational costs. However, in large-scale networks with large number of users and RRH's, using MI-SOCP is either inefficient or even intractable in some cases. Therefore, we introduce an inflation based polynomial time algorithm. We show, with numerical simulations, that our approach provides close-to-optimal solutions with much smaller amount of time compared to that needed for getting the global optimal. We also show that our approach is more efficient than the other state of the art algorithms, besides it always yields an integer feasible solution as opposed to other relaxation techniques. We also point out the essential impact of caching scheme on the trade-off between transmit power and backhaul costs. It turns out that content redundancy at the local caches, enables the RRH's to cooperate for transmit power cost reduction, but at the expense of increased backhaul cost