With the emergence of application systems which encapsulate databases and related application components, pure data integration using, for example, a federated database system is not possible anymore. Instead, access via predefined functions is the only way to get data from an application system. As a result, retrieval of such heterogeneous and encapsulated data sources needs the combination of generic query as well as predefined function access. In this paper, we present a middleware approach supporting such a novel and extended kind of integration. Starting with the overall architecture, we explain the functionality and cooperation of its core components: a federated database system and a workflow management system connected via a wrapper. Afterwards, we concentrate on essential aspects of query processing across these heterogeneous components focusing on the impact of the functions included. We discuss the operations the wrapper should provide in order to extend the workflow system's native functionality. In addition to selection and projection, these operations could include aggregation and the support of subqueries. Moreover, we point out modifications to the traditional cost model needed to consider the cost estimates for the function calls as well.

With the emergence of so-called application systems which encapsulate databases and their applications, pure data integration using, for example, a federated database system is not possible anymore. Instead, access via predefined functions is the only way to get data from an application system. As a result, the combination of generic query as well as predefined function access is needed in order to integrate heterogeneous data sources. The authors focus on a middleware approach supporting this novel and extended kind of integration. Starting with the overall architecture, we explain the functionality and cooperation of its core components: a federated database system (FDBS) and a workflow management system (WfMS). Afterwards, we concentrate on the key problems of function integration by discussing query execution planning, precedence control of function execution, and parameter handling. In this context, we develop a lightweight description language based on XML for the global-to-local mapping of functions. In addition, we consider some important aspects of the execution model, focusing on the interaction of the FDBS and the WfMS as well as the support of distributed transactions.