Max-min fairness driven multicast sparse beamforming for cache-enabled Cloud RAN

Abstract Cloud radio access network (C-RAN) is able to boost network capacity through centralized scheduling of multiple distributed base stations (BSs). Yet it also places serious burdens on backhaul links. Caching and multicast are two enabling candidates to alleviate backhaul cost and achieve efficient content delivery. In this paper, we consider a cache-enabled multicast C-RAN, where multiple BSs cooperatively serve multiple user groups. Each BS is equipped with local storage and is connected to the central processor (CP) via a backhaul link. Then, the beamforming vectors and dynamic BS clustering is carefully designed to maximize fairness among users. To achieve such a goal, a semidefinite relaxation (SDR)-based iterative difference-of-two-convex-function (D.C.) algorithm is proposed by using semidefinite programming (SDP) and smoothed l 0 -norm approximation approaches. After that, to reduce computational complexity, a two-tier convex quadratic-based alternating (TCQA) algorithm is devised by decoupling multi-ratio fractional constraints. In this algorithm, the outer problem is handled in closed-form expressions while the inner one is solved by using D.C. programming. Convergence and complexity of the two proposed algorithms are analyzed. Finally, extensive simulation results demonstrate that our proposed TCQA algorithm significantly outperforms the SDR-based iterative D.C. algorithm.