The first-come, first-served strategy is generally applied to implement customer premises equipment (CPE) wide area network management protocol, but may work poorly under overloaded conditions. Namely, many sessions established between CPEs and the corresponding auto configuration server (ACS) must be kept waiting. Therefore, a dynamic rescheduling and self-adaptive estimation (DRASE) approach is proposed in this letter to increase the throughput of an ACS with an acceptable drop rate. Specifically, the order of rescheduling is determined by the estimated processing time (EPT), the remaining lifetime, and the waiting time of each session. Once the session is completed, a self-adaptive technique is adopted to update the EPT to increase the effectiveness of rescheduling. Experimental studies show that the throughput of an ACS using DRASE at peak times reaches 1.129 times of the original value, and other evaluated performance metrics are also significantly improved.