With the proliferation of delay-tolerant applications (email, P2P networks, social networking updated, file transfer) for digital devices, in this paper, we propose a wireless transmission scheme for delay-tolerant message over Cognitive Radio (CR) networks. Since it becomes feasible to estimate the PU connections for the Secondary Users (SU) by contacting a trusted database containing the information of Primary Users (PU) traffic, our methods explicitly consider the distribution of SU traffic loads for the sake of load-balancing based on historical statistics of PU traffic. By modeling a M/M/K/L queuing system, we analyze the performance of the SUs competition under the traffic blocking thresholds and queuing delay. In this manner, we derive the number of SU connection can be accommodated simultaneously, thus optimize the frequency channel utilization. We simulate multiple message and users competing for the spectrum white space in CR networks and show that our optimal transmission scheme significantly reduces the energy cost under different traffic blocking constraints.