| Printed circuit boards are customized devices that are frequently used in electronic products. To provide an appropriate choice for a proposed new product, it requires integrated design, supplier and manufacturing planning for modular components where suppliers and manufacturing resources are network distributed. The goal of this work is to develop a comprehensive model to support such enterprise-level decision-making in network-based scalable systems, integrating the characteristics of evolutionary algorithms with the characteristics of network communication.;A great amount of research work has been focused on the self-similar nature in network traffic, which implies heavy-tailed data transmission at the individual level. In order to solve the models analytically, we propose a novel hyperexponential transition to characterize heavy-tailed network traffic, which extends the modeling power of Petri nets. Further, a hyperexponential-based network traffic model is introduced and validated by simulation.;Although a wide variety of evolutionary algorithms implement their evolutionary computation in different manners, most of them employ a similar structure at the highest level. Accordingly we transform the evolutionary operations of an evolutionary algorithm into a sequence of timed activities and apply hyperexponential Petri nets to represent these activities. Models of network-based decision systems are developed by integrating network traffic models with evolutionary algorithm models.;Performance analysis can be achieved by two directions: first, analyze and reconfigure the network connection for a specific algorithm, and second, given the network configuration, choose or redesign a suitable algorithm. Our results show that the steady state for a typical network-based system is rarely reached, suggesting that transient analysis is more important for this class of distributed decision systems. We also emphasize the failure of traditional Poisson models in the heavy-tailed network environment. At last, a number of techniques are proposed to improve the computational efficiency of network-based decision systems. |
