Workflow Management Systems
For several years the area of workflow management systems was a central research area of the group. Our main contributions are:
- Architecture of workflow systems: We developed and experimented with several architectures and finally developed the prototype workflow system “Panta Rhei”. The architecture is remarkable for its very lean kernel, which allows great flexibility in workflow processing. Our prototype is integrated into a database management system in a way that the transaction mechanisms of the DBMS can be employed for recovery of workflow processes in case of failures. Another feature which was widely recognized was the integration of the prototype with web technologies (already in 1995) which for an example allowed the participation in the processing of a workflow to anybody by simply assuming the availability of a web browser.
We introduced a highly flexible workflow description language and implemented it in our workflow engine. The distinctive features of this language were the combination of flow control structures with a form metaphor which resulted in a very high expressiveness while keeping the structures simple.
In a joint project with industry, our prototype workflow system was transferred into an industrial prototype. In a spin-off enterprise (Groiss Informatik) an industrial strength product (Alta Vista Works, later @enterprise) was developed and successfully installed in many customers sites in Austria, Germany, and Switzerland primarily in banks, governmental agencies, post offices, and hospitals. The system currently supports several thousand users.
- Workflow Transactions and Recovery: We proposed a method for automatic and semi-automatic treatment of errors and exceptions in the execution of workflow processes which built the basis for several research accomplishments by several other groups worldwide. The classification of errors and exceptions (in particular the distinction between expected and unexpected exceptions and the consequences) is frequently referenced in the literature. Core of our approach is the automatic generation of a series of valid workflow executions which combine a backwards directed recovery from failures and exceptions with a process reversal into a forwards directed pursuit of the workflow goals as early as possible.
A specific variant of this methodology was implemented for the industrial workflow management system CSE workflow, a groupware oriented workflow management system (at that time the second largest in terms of installations in Europe).
- Time Management in Workflow Systems: We developed a technique to compute time plans (internal deadlines) for all the activities of a workflow process for controlling the temporal conditions of the execution of workflows given overall deadlines. We extended this technique also for explicit temporal constraints defining temporal dependencies between the execution times of workflow activities. We developed effective procedures and algorithms to compute time plans for workflow executions satisfying all constraints. Research on this topic is continuing, in particular with the goal to compute the urgency of activities to support users in scheduling their work-load.
- Service-based Interorginzational Processes: The architecture we proposed also supports the execution of workflows spanning several organizations and the cooperation of workflow systems for inter-organizational workflows which we frequently find in e-commerce and e-government. This concept of „world wide workflows” was also implemented in a prototype. Currently we extend it with a module for the cooperation of workflow systems by the exchange of XML documents.
We develop a system for the adequate representation of changes in dimension data of multi-dimensional data warehouse models which enables to correctly represent change histories of dimension data , identify temporal versions based on these changes and permit the correct evaluation of OLAP queries in spite if these changes. A project for implementing this approach in an industrial prototype is currently starting.
In the area of deductive databases we made contributions to the development of language constructs for expressing recursive queries and for optimizing such queries. The concept of generalized transitive closures brought high expressivity without loosing the ability for automatic query optimization. The recursive constructs in the commercial database system DB2 are very similar to the concepts we developed and proposed. Rather exclusive was the extension of recursive queries to multi-set relations which are the formal basis for the tables in SQL databases.
Modelling of the Dynamics of Information Systems
For several years we contributed to the development of static and dynamic models for information systems. The “Behaviour Integrated Entity Relationship Model” was among the very first approaches to combine static and dynamic models. In our newest results, we proposed techniques for manipulating dynamic models in a similar way as static models to perform model transformations for improving the quality of dynamic models. Furthermore, we presented methods for integrating dynamic models extending the view integration approach of conceptual modeling form static models also to dynamic models. This technique was then also taken over for manipulating workflow models.
Views on Object-Oriented Databases
The concept of views (external models in the terminology of ANSI/SPARC) permits logical data independence and thus contributes enormously to the reduction of complexity as well in the development of databases as in the maintenance of database systems. We extended this concept with all wanted features to object-oriented databases and introduced a solution for external models which can be integrated completely in contra-variant inheritance hierarchies and guarantees method steadiness.