An optimal release policy for software testing process

In this paper, we discuss the dynamic release problem in software testing processes. If we stop testing too early, there may be too many defects in the software, resulting in too many failures during operation and leading to significant losses due to the failure penalty or user dissatisfaction. If we spend too much time in testing, there may be a high testing cost. Therefore, there is a tradeoff between software testing and releasing. The release time should be dynamically determined by the testing process. The more defects have been detected and removed, the less time will be used for further testing. A continuous time Markov process is proposed to model the testing process. By formulating with dynamic programming we obtain the Hamilton- Jacobi-Bellman equation of the optimal cost function, and derive the threshold structure of the optimal policy. Furthermore, the dynamic optimal release policy is compared with the static optimal release policy by numerical examples, showing that dynamic policy may outperforms static policy very much in some situations.