Considering spatial and temporal features of urban storms, a proper surface runoff coefficient is selected to measure the quantity of flowing water, which will be collected by inlets. The sequence of drainage consists of rainfall collecting, conveying and discharge through inlets, pipes and pumps respectively, referred to a cascade connection queuing system. According to their importance, an urban district is divided into several areas whose rainfall should be drained out within a restricted time determined by their importance. The time, as well as the cost of building and the operation of the urban drainage system (UDS), is set as constraints of a biobjective optimization model for the pump location problem. Then we develop an optimization algorithm, genetic algorithm integrated with Tabu algorithm, to solve this complex nonlinear allocation problem. A practical case study reveals that multiple cascaded connection queuing method can simulate the UDS well, and the quantity and location of pump stations, as well as storm recurrence period, plays an important role in designing an UDS.