Mobile network operators usually consider power consumption and Grade of Service (GoS) as two important aspects in the design and planning of modern cellular networks. Base station (BS) sleeping is an effective approach to reduce the power consumption of the network, by switching some of the BSs to a low-power “sleep mode” during off-peak traffic hours. In this paper, we model each BS with sleeping mechanism as an M/G/1/K queue with vacations, and the entire cellular network as a network of such queues, to incorporate practical factors in BS sleeping, such as close-down and startup periods and additional power consumption for activating a sleeping BS. We investigate the power consumption and GoS under three BS sleeping schemes: (1) the isolated scheme, in which each BS switches between active and sleep modes based on its own realtime traffic load, (2) the cooperative scheme, in which selective BSs are switched to long-term sleep and traffic is allowed to overflow from sleeping BSs to nearby active BSs, and (3) the hybrid scheme, in which some BSs are switched to long-term sleep and other BSs switch modes according to their real-time traffic load. A robust, scalable and computationally efficient analytical method is proposed to evaluate GoS metrics, including mean delay and blocking probability, and power consumption under each scheme. We validate the accuracy of the proposed method, demonstrate the trade-off among power consumption, blocking probability and mean delay, and compare the performance of the three schemes via extensive and statistically reliable numerical experiments.
Read full abstract