Reducing Energy Consumption of Radio Access Networks

Radio access networks (RANs) have become one of the largest energy consumers of communication technology [LLH+13] and their energy consumption is predicted to increase [FFMB11]. To reduce the energy consumption of RANs different techniques have been proposed. One of the most promising techniques is the use of a low-power sleep mode. However, a sleep mode can also reduce the performance. In this dissertation, I quantify how much energy can be conserved with a sleep mode and which negative effects it has on the performance of RANs. Additionally, I analyze how a sleep mode can be enabled more often and how the performance can be kept high. First, I quantify the effect of power-cycle durations on energy consumption and latency in an abstract queuing system. This results in a trade-off between energy consumption and latency for a single base station (BS). Second, I show that considering a network as a whole (instead of each BS individually) allows the energy consumption to be reduced even further. After these analyses, which are not specific for RANs, I study RANs for the rest of the dissertation. RANs need to both detect and execute the requests of users. Because detection and execution of requests have different requirements, I analyze them independently. I quantify how the number of active BSs can be reduced if the detection ranges of BSs are increased by cooperative transmissions. Next, I analyze how more BSs can be deactivated if the remaining active BSs cooperate to transmit data to the users. However, in addition to increasing the range, cooperative transmissions also radiate more power. This results in higher interference for other users which slows their transmissions down and, thus, increases energy consumption. Therefore, I describe how the radiated power of cooperative transmissions can be reduced if instantaneous channel knowledge is available. Because the implementation in real hardware is impractical for demonstration purposes, I show the results of a simulation that incorporates all effects I studied analytically earlier. In conclusion, I show that a sleep mode can reduce the energy consumption of RANs if applied correctly. To apply a sleep mode correctly, it is necessary to consider power-cycle durations, power profiles, and the interaction of BSs. When this knowledge is combined the energy consumption of RANs can be reduced with only a slight loss of performance. Because this results in a trade-off between energy consumption and performance, each RAN operator has to decide which trade-off is preferred.