摘要：If a software product with a significant numberof defects is released too early to users, the softwaremanufacturer will incur post-release costs of fixing the faults.If a product is released too late, the additional developmentcost and the risk of missing a market window could besubstantial. Software Reliability Growth Models (SRGMs)can capture the quantitative aspects of testing and areused to estimate software release time. From a cost-benefitviewpoint, SRGMs aid developers to decide the optimalrelease time of the software product by providing effectiveapproaches to minimising the expected total software systemcost. This paper helps answer the question of when to stoptesting a software product by presenting the perspectivesfrom a study of cost models. The study focuses on aspectsof the relationship between development cost and scheduledelivery of the software product and the total software costincluding the risk costs, such as the penalty cost incurred dueto late delivery of software product and the cost of fixing afault during the warranty period.We also investigate varioussoftware release policies, for example, policies based on thedual constraints of cost and reliability.
