The effects of a primary user's (PU) multiple random arrivals and departures on the throughput of the sensing-based spectrum sharing cognitive radio networks are investigated in this paper. We give a comprehensive study on the probability that the PU change its state n times during the whole frame time, and propose a new model to describe the PU traffic. Based on this proposed model, an optimal power allocation strategy with four-level is presented, to maximize the achievable throughput of the SU. Simulation results show that the throughput of the SU can be improved considerably by our scheme.