Modeling exceptional behaviors in commercial workflow management systems

Workflow management systems (WfMSs) are most applicable when processes follow a predictable and repeatable behavior. However exceptional situations may occur in any process. An important class of exceptions is that of expected exceptions, i.e., of those exceptional situations that can be foreseen at design time, are part of the process semantics, but cause deviations from the normal process execution. Expected exceptions often are asynchronous with respect to the normal flow. Recently, a few research prototypes have addressed the problem of supporting exceptional behaviors, while commercial WfMSs do not offer these capabilities yet. In this paper we propose a methodology for modeling exceptions by means of activity graphs, which are the workflow specification formalism supported by most commercial systems. We present a taxonomy of expected exceptions, categorizing them according to how then can be mapped. Finally, we give methodological guidelines in order to support exception analysis and design. We exemplify the introduced concepts by showing how expected exceptions can be mapped on top of Changengine, Hewlett-Packard's WfMS.