Abstract
When we make a design of a computer network system, it is necessary to investigate the performance of the system at steady state. In the early stage of computer system design, we have many choices and use simulations for performance analysis. We can not, however, examine system performance satisfactorily, because it takes much time to analyze all of the choices for large scale systems.
In this paper, we will present a development of a performance evaluation system “MELMATE” which use a queueing network theory. MELMATE can give us the results of analysis more than ten times faster than conventional event driven simulations. Also, MELMATE is bundled with GUI (Graphical User Interface) which makes model construction and modification much more easier. That is, we can check up every possible design of the system in a very short time and select the best one comparing with the obtained performance results. The methodology of a performance evaluation system and an example of usage will be described.
