Base station (BS) sleeping is an effecting way to improve the energy-efficiency of cellular networks. Considering BS sleep mode operation under different scenarios, we focus on three wake-up policies: single vacation (SV) policy, multiple vacation (MV) policy and N policy. A hysteresis time is also considered to avoid frequent BS mode-changing operation. By modeling the systems as M/G/1 vacation queues, we derive two performance measures of interest, expected system response time and energy consumption per bit. The impacts of sleep mode operation parameters and setup time are also studied. In order to determine the optimal parameter settings, which allows for a flexible tradeoff between energy-efficiency and mean delay, a two-step optimization method is proposed. We numerically analyze and investigate the energy-delay tradeoff for different policies, and find that both the MV policy and the N policy have better performance than the SV policy if the sniffing cost is not considered. However, these advantages diminish when the
sniffing cost increases.