Cloud computing provides virtualized, dynamically-scalable computing power. At the same time, reduction of cost is also considered as an important advantage of cloud computing. Data integration can notably benefit from cloud computing because integrating data is usually an expensive task. However, existing optimization techniques pay less attention on the fact that different execution plans of the same data integration application generate different usage costs while cloud computing provides good enough performance, so this paper introduces the cost optimization of cloud-based data integration system. The data integration system's data service layer facilitates accessing and composing information from a range of enterprise data sources through data service composition. In addition, two task scheduling algorithms for parallel part and non-parallel part are proposed to minimize the usage cost required to complete the execution of composite data service when computational capability provided by cloud computing is charged. Both of the two can obtain optimal plans in polynomial time. Experiments with the system indicate that our algorithms can lead to significant cost saving over more straightforward techniques.