Capabilities-based query decomposition and optimization in heterogeneous data integration system

A heterogeneous data integration system can integrate a broad range of data sources that provide different levels of query capabilities. Some data sources can only support simple scanning, such as file systems; some data sources can support query by using boolean expressions, such as text retrieval systems; some data sources can even support complex operations such as joining in relational database systems. The declarative approach, which uses some languages to express query capabilities of various data sources, is difficult to describe some particular limits for query processing, while the encapsulation approach leads to very complex wrappers. To simplify the design of wrappers for data sources with limited capabilities, a kind of approach, named compound approach, which can take advantage of all powers of various data sources, is proposed for query decomposition and optimization. Different from the declarative approach and the encapsulation approach proposed formerly, the compound approach only needs brief declarations of basic query powers, but leaves the minute details of the capabilities and restrictions of the underlying data sources encapsulated in their wrappers. The concept of capability vector is proposed for describing basic query powers of data sources. Furthermore, with subsuming queries and filters of OIM algebra operations, the way for processing of queries basically supported in wrappers is given.