Abstract
The purpose of this paper is to develop a new method for information system sizing and selection based on a hybrid mixture of simulation and genetic algorithm, and to show its cost-effectiveness by applying it to a real world problem. To serve this purpose, we propose an operational model which identifies a set of system alternatives using simulation, and determines the optimal one using genetic algorithm. Specifically, with simulation, we generate probability distributions describing real data gathered from actual system, which can overcome the major weakness of the existing methodology that normally employs point estimates of the actual data and constant correction factors without theoretical rationale. We next search for the optimal combination of H/W, the number of CPUs, and S/W, which meets both of our business goals of incurring low TCO(total cost of ownership) and maintaining a good level of transaction processing performance. Experimental result shows the proposed method in this paper saves the cost while it preserves the system's capacity within allowable performance range.