This study proposes an energy-saving-centric downlink scheduling scheme to support efficient power utilization and to satisfy the QoS requirements. The base station considers the queue lengths of mobile stations with real-time and non-real-time connections and considers their QoS requirements to determine the sleeping parameters when the mobile stations issue sleep requests. The proposed scheme appropriately reschedules the sleep-requesting mobile station to transmit its queued packets for optimal power-saving efficiency. The QoS requirement is considered as the constraint during traffic rescheduling. The treatment of real-time connections generally requires a trade-off of the delay requirement and the longer sleep window, and the non-real-time connections must concern the packet drop and minimum data rates when performing the energy-centric scheduling. Two rescheduling algorithms, that is, whole and partial reschedules, are proposed and analyzed in this paper. The whole-reschedule scheme provides improved energy-saving performance at the cost of tolerable longer delay and computing complexity when compared with the partial-reschedule scheme. Our simulation results indicated that both schemes not only guarantee the desired QoS but also achieve superior energy-saving efficiency to that of traditional scheduling.