[Back]

C. Mayr:
"Managing and Modeling Persistent Data Access in Process-Driven SOAs";
Supervisor, Reviewer: S. Dustdar, U. Zdun; Institut für Informationssysteme, AB Verteilte Systeme, 2012; oral examination: 2012-07-19.

In process-driven service oriented environments, process activities in a business process can invoke services to fulfill certain tasks. A service can e.g. invoke other services, perform business logic, or read and write data from persistent data storages. In particular, in long-running business processes, which require human interaction with users, process activities often need to read or write persistent data.
Commonly this persistent data access is encapsulated by a special type of services - the Data Access Services (DAS). With these DAS, business processes can access persistent data both technology-independently and database-neutrally.
Unfortunately, these DAS are not sufficiently integrated into process-oriented environments, yet. Accordingly, the relationships between the different concerns of a process such as the process activities, the DAS, the underlying object-relational mappings (ORM), and the persistent data storages are not well-defined, yet. Moreover, different stakeholders have to be able to focus on these different concerns of the process selectively. For example, in order to prevent, detect, and solve structural problems in business processes such as deadlocks, stakeholders need to have a tailored views into the persistent data access of the process. When the number of process activities grows along with the number of DAS, finding certain DAS of a process can become an impossible task for the stakeholders.
In this thesis we focus on better integrating persistent data access into process-driven service-oriented environments. In order to achieve this, firstly, we introduce the viewbased data modeling framework (VbDMF) designed to specify persistent data access of processes. Based on VbDMF, we illustrate the concept of persistent data access flows used to prevent, detect, and solve structural problems in processes. In order to be able to better maintain and reuse persistent data access, we present an integral architectural approach to manage our VbDMF models and model instances: The viewbased model-driven data access architecture (VMDA). Along with our concepts, we illustrate the applicability by a number of industrial case studies. Furthermore, we prove the complexity and correctness of the presented algorithms. Finally, we quantitatively show, that the VMDA returns with acceptable response times.