For the series production system with waiting time constraints, breakdowns occur stochastically at the upward and downward stations. The big buffer size can cause high rework cost, and the small buffer size can cause high shortage cost. To ensure the quality of products and to reduce the operation cost effectively, an optimization model is proposed. Firstly, based on the characteristic of waiting time constraints, the preventive maintenance cost, rework cost and shortage cost are analyzed. Then, a model is established with an objective to minimize the total cost per unit time. The optimal buffer size is derived. Finally, the result of neumerical computation shows that, the optimal buffer size can be determined by using the proposed model, which can reduce the operating cost effectively.