<title>Development and evaluation of PACS system simulator</title>

ABSTRACT It is necessary to estimate system performance to ensure cost-effective system design. In addition, discrepancies betweenestimated performance and system performance must be established as a guidelines for the use of the simulator in advance.We have developed and evaluated a PACS system simulator,'UNSUI/P". UNSUI/P is based on discrete event simulator,"UNSUF', developed by Toshiba Research and Development Center.UNSUI/P was evaluated by simulating a number of PACS and comparing the results against actual system performance.Evaluations for a single component and for the entire system were carried out. Discrepancies between simulation results andactual system performance ranged from 30% to 50%for the response time in the first evaluation. In the second evaluation,simulation of the delay time for image transfer from a modality to a viewing workstation was performed and discrepancies inthe delay time appear to be about 15% to 40%. 1. INTRODUCTION The design for PACS must be adapted to each specific installation site. For such design, it is necessary to estimate theperformance based on the number of components and the kind of network employed. A cost-effective system is established byoptimizing the trade-off between system performance and the cost of system components.A numerical method for queuing networks has been used to estimate system performance. It has good availability;however, some statistical assumptions must be made to construct a system model. Due to such assumptions, largediscrepancies are often seen in the results for complex system performance estimation. This method is also unable to analyzesystem overload behavior and is unable to estimate overload performance. On the other hand, a discrete event simulator haslooser constraints for model description than the numerical method. It is able to describe detailed models and is suitable forsimulating specific sites. In addition, it is able to analyze system overload behavior.First, we describe the configuration of a discrete event simulator, UNSUI,P, aiil report two kinds of results for PACSperformance estimation using the simulator. Then, the discrepancies between the simulation results and actual systemperformance for evaluation of UNSUI/P will be described.642 /SPIE Vol. 2165