Approximate Performance Analysis of Web Services Flow Using Stochastic Petri Net

Abstract

Web services provide a language-neutral, loosely-coupled, and platform-independent way for linking applications within organizations or enterprises across the Internet. In such a scenario, quantitative characteristics such as service execution throughput should be evaluated to measure the system performance. Usually, the first step is to define an abstract workflow model, for example Stochastic Petri Net Models. However, large system always raises the problem of state explosion. In this paper, we discuss a set of simplification rules for five basic structures of web service flow: sequential, parallel, conditional, loop and mutex. Our approach can effectively reduce the state space and is applied to the performance analysis of a web service flow management system. © Springer-Verlag 2004.