Abstract: With the rapid development of Internet and WWW, integration of distributed and diverse data sources becomes a more and more important issue However, it is a difficult problem due to heterogeneous data structure and capabilities of these data sources One possible solution is to construct query plans according to the data source capabilities which are described as metadata A variety of methods serve this purpose due to different application requirements, two typical methods are studied here, one tries to construct simple but efficient query plan as quickly as possible, the other aims on getting as many semantically correct query answers as possible by exploiting semantic relationships among data sources These semantic relationships are expressed as mappings between the logical data space and each concrete data source Experiments are carried out to analyze and demonstrate the efficiency and effectiveness of these methods In the same time, complicated query conditions are also considered in order to assign appropriate query condition to each data source