Performance Evaluation of Workflows Using Fluid Flow Approximation of Discrete Sets and Probability Distributions

How to make information systems safe, dependable and trustworthy is one of central concerns in the development of e-Society and e-Government. In this paper, we focus on the problem of guaranteeing performance of workflow-based information systems. Performance evaluation by traditional stochastic approaches such as queuing models and stochastic Petri nets is costly, and is hard to be applied to workflow systems in which many instances of flows run simultaneously. We develop a novel method for deriving performance indices from hybrid- system models in which each discrete set is approximated by a continuous region on a real-valued vector space and the variance of each probability distribution is represented by a real-valued interval. To prevent approximated ranges from divergence, we use place invariants as constraints on the state space. Paper review processes in an academic journal are used as a case study.