Quality-driven optimized resource allocation

Abstract

The assurance of a good software product quality necessitates a managed software process. Periodic product evaluation (inspection and testing) should be executed during the development process in order to simultaneously guarantee the timeliness and quality aspects of the development workflow. A faithful prediction of the efforts needed forms the basis of a project management (PM) in order to perform a proper human resource allocation to the different development and QA activities. However, even robust resource demand and quality estimation tools, like COCOMO II and COQUALMO do not cover the timeliness point of view sufficiently due to their static nature. Correspondingly, continuous quality monitoring and quality driven supervisory control of the development process became vital aspects in PM. A well-established complementary approach uses the Weibull model to describe the dynamics of the development and QA process by a mathematical model based on the observations gained during the development process. Supervisory PM control has to concentrate development and QA resources to eliminate quality bottlenecks, as different parts (modules) of the product under development may reveal different defect density levels. Nevertheless, traditional heuristic quality management is unable to perform optimal resource allocation in the case of complex target programs. This paper presents a model-based quality-driven optimized resource allocation method. It combines the COQUALMO model as early quality predictor and empirical knowledge formulated by a Weibull model gained by the continuous monitoring of the QA process flow. An exact mathematical optimization technique is used for human resource, like tester allocation.