Modeling the Behavior of Dispatching Rules in Workflow Systems: A Statistical Approach

摘要

Using scheduling techniques to reorder work in workflow systems can improve the performance of the business processes enacted by these systems. But this research area is still under-explored. This paper presents preliminary results on ongoing research on the modeling of the behavior of scheduling techniques in workflow systems using statistical analysis. It discusses the the motivations for this approach and presents a first workflow scenario to be studied. Simulations of the use of dispatching rules to minimize the percentage of late cases are performed. The results are analyzed with statistical regression techniques to estimate functions that describe the observed data and compose a model to the behavior of the scheduling techniques. The model is evaluated against new data that was not present in the first observed data and the results show that it is feasible and could be used to assist systems administrators in applying scheduling techniques to workflow systems.