Workload-aware Business Process Simulation with Statistical Service Analysis and Timed Petri Net

To adapt systems in response to new business requirements with SLA, business process performance simulation will play a key role. This paper proposes a performance model for business process that comprises control flow structure (such as loop) and parallel message execution under shared resources. A network queuing model is inappropriate to model dynamic workload in the business process execution since messages are processed in parallel consuming shared resources. In our previous work, we have proposed a statistical service model to estimate execution time based on workload overlap analysis. In this paper, we extend this approach to model a business process with control flow. We integrate a Timed Petri Net model and our statistical models for services and process engines. We have implemented a prototype system for supply chain scenario. Experiments with this prototype show high accuracy of performance simulation.