A Numerical Model of the Web-Based Clinical Image Reference System and System Performance Estimation from the Model

Abstract

 Performance estimation procedure for the clinical image reference system with World Wide Web architecture is presented. Here, numerical modeling and simulation methods are employed. In the reference system, simultaneous requests for images initiated by each user might cause performance degradation. Both parallel processing and competitive usage of system resources such as hard disks and network are described exactly in the model, and implemented using in Java. Reliability of this model is confirmed by an experimental network system with 32 workstations. The time during fetching and displaying an image from the simulation agrees well to the experimental values, except in extremely heavy loads. An effective data transfer rate is defined here as the product of average data transfer rate derived from the simulation and number of workstations. This index is convenient to estimate the maximum number of workstations for the given system circumstances. For some cases of operations in the clinic with average image request interval of 5 minutes or 3 minutes, the relationship between maximum number of workstations and number of images at a request is estimated. We can conclude that this performance estimation procedure is quite useful for the design of the clinical image reference system.