Discrete Optimal Testing/Maintenance Policy in a Software Development Project

In this paper, we focus on the optimal software testing/maintenance problem considered in Rinsaka and Dohi (2004), and develop a stochastic model in discrete operation circumstance, under the difference between the software testing environment and the operational environment can be characterized by an environment parameter. More precisely, the total expected software cost is formulated via the discrete NHPP type of software reliability models. In the special case with the geometric fault-detection time distribution, we derive analytically the optimal testing period (release time) which minimizes the total expected software cost under a milder condition. We call the time length to complete the operational maintenance after the release a planned maintenance limit, and also derive the optimal planned maintenance limit which minimizes the total expected software cost. In numerical examples with real data, we calculate numerically the joint optimal testing/maintenance policy, combined by testing period and planned maintenance limit.