Supporting enterprise-wide or even cross-organizational business processes is a characteristic challenge for any workflow management system (WfMS). Scalability at the presence of high loads as well as the capability to dynamically modify running workflow (WF) instances (e.g., to cope with exceptional situations) are essential requirements in this context. Should the latter one, in particular, not be met, the WfMS will not have the necessary flexibility to cover the wide range of process-oriented applications deployed in many organizations. Scalability and flexibility have, for the most part, been treated separately in the relevant literature thus far. Even though they are basic needs for a WfMS, the requirements related with them are totally different. To achieve satisfactory scalability, on the one hand, the system needs to be designed such that a workflow instance can be controlled by several WF servers that are as independent from each other as possible. Yet dynamic WF modifications, on the other hand, necessitate a (logical) central control instance which knows the current and global state of a WF instance. For the first time, this paper presents methods which allow ad-hoc modifications (e.g., to insert, delete, or shift steps) to be performed in a distributed WfMS; i.e., in a WfMS with partitioned WF execution graphs and distributed WF control. It is especially noteworthy that the system succeeds in realizing the full functionality as given in the central case while, at the same time, achieving extremely favorable behavior with respect to communication costs.
|Name||CTIT Technical Report Series|
|Publisher||Centre for Telematics and Information Technology, University of Twente|