Business process management (BPM) is an increasingly challenging aspect of the enterprise. Middleware support for BPM, as provided by, for example, Oracle, Biztalk and the recent Windows Workflow Framework, has met some challenges with respect to performance and maintenance of workflow.
The central challenge to BPM is complexity: business processes are becoming widely distributed, interoperating across a range of inter and intra organizational vocabu laries and semantics. It is important that complex business workflows are checked and analyzed for optimality and trustworthiness prior to deployment. The problem becomes worse when we consider the enterprise’s demand to regularly adapt and change processes. For example, the growth of a company, changes to the market, revaluation of tasks to minimize cost. All these factors often require reengineering or adaptation of business processes along with continuous improvement of individual activities for achieving dramatic improvements of performance critical parameters such as quality (of a product or service), cost, and speed . Reengineering of a complex workflow implementation is dangerous, due to existing dependencies between tasks.
Formal methods can assist in meeting the challenge of complexity, as their mathematical basis can assist in analysing and refining a system specification. However, complex systems often involve a number of different as pects that entail separate kinds of analysis and, consequently, the use of a number of different formal methods.
Petri nets are a formal method that has successfully assisted in workflow design and analysis. While Petri nets are good for expressing the dynamics of a workflow, the representation of data as tokens do not provide the full depth of specification necessarily to by developers. Petri nets model the possible flow of information in a business process, but do not specify the nature of the information nor how in formation is to be manipulated during the business process. Petri net lack modeling power and mechanisms for data abstraction and refinement .
In contrast, a business process implementation within a BPM middleware requires detailed treatment of both information flow and information content. The abstraction gap is identified by Hepp and Roman in : an abstract workflow that ignores information content provides an abstract view of business processes that does not fully define the key as pects necessary for BPM implementation.
We argue that this abstraction gap can be addressed by developing semantically compatible PN models and data models from an initial business process requirements specification. We employ a Model Driven Architecture approach.
The overall framework is depicted in Fig. 1. For our purposes, we consider transformations between models of three languages, a Computation Independent Model (CIM) and two Platform Independent Models (PIMs). The CIM is aninitial model of business requirements. It describes business functionality without treating any architectural or computational aspects of the system implementation. The two PIMs describe complementary aspects of the overall structure of the system to be implemented: workflow descriptions from PN models and data and data exchange mechanisms from Event B specifications.
Download
PDF Ebook Normative Ontologies for Data-Centric Business Process Management
