Flexibility for Distributed Workflows [chapter]

Manfred Reichert, Thomas Bauer, Peter Dadam
Enterprise Resource Planning  
This chapter shows how flexibility can be realized for distributed workflows. The capability to dynamically adapt workflow instances during runtime (e.g., to add, delete or move activities) constitutes a fundamental challenge for any workflow management system (WfMS). While there has been significant research on ad-hoc workflow changes and on related correctness issues, there exists only little work on how to provide respective runtime flexibility in an enterprise-wide context as well. Here,
more » ... lability at the presence of high loads constitutes an essential requirement, often necessitating distributed (i.e., piecewise) control of a workflow instance by different workflow servers, which should be as independent from each other as possible. This chapter presents advanced concepts and techniques for enabling ad-hoc workflow changes in a distributed WfMS as well. Our focus is on minimizing the communication costs among workflow servers, while ensuring a correct execution behavior as well as correctness of ad-hoc workflow changes at any time. To deal with these requirements it is crucial to identify the WF servers of the distributed WfMS to be involved in the synchronization of an ad-hoc change. Most likely we have to consider those WF servers currently executing the respective WF instance. These active servers need to know the schema and state of a changed WF instance in order to correctly control its execution afterwards. We need an efficient approach for determining the set of active servers controlling a particular WF instance. This must be possible without a substantial expense of communication efforts. In addition, we have to decide whether, when and how a changed WF instance schema has to be transmitted to other WF servers. As essential requirement the amount of communication should not exceed acceptable limits. This chapter is structured as follows: We first give background information needed for the further understanding and we introduce basic issues related to distributed WF execution as accomplished in the ADEPT approach. Following this, we first describe how ad-hoc instance changes can be performed in the distributed variant of the ADEPT WfMS. Then we show how individually modified WF instances can be efficiently executed in such distributed WfMS. Finally, we describe our proof-ofconcept prototype and discuss related work. The chapter concludes with a summary and outlook.
doi:10.4018/978-1-4666-4153-2.ch072 fatcat:vspidsmjmjdyrkrvatacl7f66y