Massively parallel simulation of business process models

Business Process Reengineering has emerged as a modeling and analysis methodology for making the organization of an enterprise more efficient. For the purpose of reengineering, a business process model (BPM) has to be developed, the evaluation of which in almost any case demands simulation. Even for enterprises of moderate size, the BPMs tend to become prohibitively complex, precluding a sequential simulation as an efficient evaluation method. In this paper, timed Petri Nets are used for the specification of BPMs, for which a massively parallel simulation engine was developed and implemented on the CS-2 multiprocessor. Executing a BPM of realistic size on the CS-2 considerably reduces the elapsed simulation time. Specifically, parallel simulations of large BPMs reveal a 300 fold acceleration when executing on 32 processors.