Understanding Energy Consumption of Cloud Radio Access Networks: an Experimental Study

Abstract

Cloud Radio Access Network (C-RAN) is rising as an attractive solution for the operators to cope with the ever-increasing user demand in a cost-efficient way. C-RAN’s architecture consists of (i) Distributed Units (DU) located at the remote sites along with RF processing units, (ii) the Central Unit (CU) consisting of high speed programmable processors performing tasks such as mobility control, radio access network sharing, positioning, session management over a (iii) low latency, high bandwidth fronthaul link, which connects multiple DUs to the CU pool realized on a cloud platform. In traditional C-RAN, the functionalities that the BBUs and RRHs have to perform are fixed. Instead of having such a fixed set of functionalities, the concept of functional splits was introduced by 3GPP to bring forth the idea of shifting network stack functions between CUs and DUs in next generation C-RAN. In this paper, a real-time C-RAN testbed running on OpenAirInterface (OAI) software platform is used to profile the energy consumed by different functional splits configured by varying the CPU clock frequency and channel bandwidth. It is observed that for some lower CPU clock frequencies, the energy consumption is reduced without affecting the system throughput and overall user experience. With these insights, operators can improve the energy efficiency of CRAN systems deployed.