Abstract
In this paper, a new method is proposed to evaluate the performance of concurrent systems. A concurrent system consisting of multiple processes that communicate via message passing mechanisms is modeled by a Petri net, which is in turn represented by a set of ordinary differential equations (ODEs) of a restricted type. The equations describe the system state changes, and the solutions, also called state measures, can be used for the performance analysis such as estimating response time, throughput and efficiency. This method can avoid a state explosion problem encountered by the conventional methods based on Continuous-Time Markov Chains. Its application to an IBM business system is given as an example.
Note to Practitioners-This work proposes a Petri net and ODE-based method for the system performance analysis. A concurrent system is first modeled by a Petri net. It is then converted into a continuous Petri net, which can be represented by a set of ODEs. Finally, the performance metrics can be derived from the ODE solutions that can be obtained in polynomial complexity. Thus, the proposed method can entirely avoid the state explosion problem often encountered by commonly used Markov chain based methods. This method can be applied to business systems, e-commerce, automated production systems, service computing, workflow analysis, and computer communication systems.