In dynamic spectrum access networks, the unused licensed spectrum of primary users (PU) is opened to unlicensed secondary user (SU) for improving spectrum efficiency. The problem we study in this paper is to maximize the SU throughput while guaranteeing PU performance. We design a simple time-based threshold strategy to maximize the SU throughput while guaranteeing the PU performance by learning directly from the interaction with PU environment. At a given time, the SU decides if it should transmit and if so to protect the PU performance. It is observed that such strategies perform closely to the strategy optimized based on prior knowledge about PU traffic models. Simulations are used for performance evaluation.