Temporal aspects in workflow management systems

The need for explicit time management in workflow environments has been recently identified. Introducing time management in WFMS has two aspects (a) allowing the specification and execution support of advanced temporal constraints and (b) allowing the acceleration (speed-up) of already executing workflow instances. In most existing approaches, time is handled with the same general-purpose mechanisms of the WFMS. Recently, it has been argued that workflow applications, like health care, GIS and cross-organisational applications, are constrained by temporal constraints whose consistency must be verified and have to be enforced at execution time. Yet, existing approaches focus on quantitative temporal constraints and do not consider qualitative aspects. In this thesis we incorporate a subset of interval algebra in workflow specification, develop a consistency-checking mechanism that verifies the consistency of the specification and examine different scheduling policies based on different application requirements.; The second aspect of temporal management arises in the context of e-commerce workflow applications. Electronic commerce applications can be described as activities consisting of a set of tasks that are structured based on data and control dependencies, and executed by a set of distributed agents, thus resembling workflow activities. With the rapid development of the electronic marketplace, companies have to adjust to a global, highly competitive and dynamic environment. Hence, it is critical for WFMSs to be capable of providing scheduling policies that allow accelerating e-commerce applications according to "on-line" speed-up requests. In this thesis, we formulate the problem of speed-up without acquiring any additional resources, for a multi-agent environment where each agent is responsible for speeding-up local tasks residing in its input queue. We model the concept of speed-up and of activity speed-up dependencies (i.e. dependencies that arise due to speed-up requests). We examine the applicability of our model when agents can delegate sub-activities to other agents in a nested activity environment. We present a cost model for satisfying speed-up requests by calculating the resultant slow-down to other concurrently executing activity instances. We design time-based and cost-based speed-up algorithms, that consistently and efficiently speed-up activities. We evaluate our results through simulation.