|
Abstract:
|
In designing and analyzing complex network systems, one is particularly interested in how to minimize the service cost by controlling a large number of input factors. This service cost usually corresponds to performance measures (e.g., delays and backlogs) for which analytical solutions are invalid due to the system's complexity. Therefore, one has to rely on simulation. Simulations of large systems are expensive in terms of CPU time and use of available resources. Thus, it is important to select the inputs of simulation carefully in order to minimize the cost function of interest and the required number of simulation runs (i.e., experimental trials). In this study, response surface methodology (RSM) is used to achieve these goals. We briefly summarize the steps of RSM and show it is a powerful tool for solving various network optimization problems.
|
