Enhancing Workflow Data Interaction Patterns by a Transaction Model

Todays process-aware information systems (PAIS) provide little support for explicit specification of transactional aspects. PAIS have to integrate events and data from various external sources as workflow relevant data. Furthermore, it should be aware of changes made externally and write consistently back data used and altered to external sources. To avoid inconsistencies within redundantly maintained data, transactional aspects within process and data perspective have to be supported. We present a layered architecture which overcomes most of these problems by extending a workflow management system (YAWL) with facilities to access external data sources, to associate the control flow perspective with transactional properties like isolation, serializability and recovery. To ensure a better data integrity, we define synchronization strategies and integrity constraints beyond single objects and tasks. Furthermore, we integrate transactional and non-transactional sources to offer better data security, data persistence and data recovery within our workflow model.