Design of experiments in software quality improvement
George J. Besseris · International Journal of Quality Engineering and Technology · 2009
The software industry and information technology (IT) research laboratories incorporate quality management methods to their everyday operations, expecting to make their products more robust to customer demands thus sustaining their market penetration against their competitors. In this work, we present a simple methodology that may assist practitioners in the field of continuous improvement in IT product development and software quality assurance to prioritise those code quality parameters that require the most attention in order to optimise run-time critical responses. Design of experiments (DOE) is employed according to Taguchi's quality philosophy to show how to discern 'cause-and-effect' relationships among code performance metrics and code oriented quality factors. Through a case study in a typical heat-flow simulations application such as often encountered in mainstream industrial operations, code quality parameters are selected to be the node count, mesh size, the numerical tolerance, the count of iterations and the solver constant.