Software Production Process for Safety Critical Software
Norman F. Schneidewind · Journal of Aerospace Computing Information and Communication · 2008
A software production model is developed that integrates process and product and is designed to identify bottlenecks in the production process. Choke points can occur either from process deficiencies or failure to identify and correct extant software defects. NASA Goddard Space Flight Center satellite defect data is used in the analysis because the primary aim is to apply the model to safety critical software. The model applies feedback control to correct anomalies in process and product that may occur. Both predictions and actual defect data are used to identify process and product behavior that do not meet expectations. I. Introduction A CCORDING to, 1 software systems come and go through a series of passages that account for their inception, initial development, productive operation, upkeep, and retirement from one generation to another. We focus on the development and production phases and introduce something different in this research compared with articles about software development process that correctly relate the effectiveness of the process to the quality of the product. We go a step further and introduce the concept of the reliability of the process, both current and future. We perform this analysis by using software defect data from the NASA Goddard Space Flight Center satellite project known as JM1. These data are shown in Table 1. Our goal, recognizing the importance of integrating process and product, 2 is to identify production bottlenecks in the process that cause product delivery to be delayed. In addition, we identify deficiencies in the defect removal process that can lead to product unreliability and delivery schedule delay. The principal features of the production system are defined using process flow charts illustrated in Fig. 1 for a software production and quality control operation. The flow of software modules is from left to right. Each phase has a processing time measured in days and a defect rate. Our model assumes that defects introduced by a process step may pass through downstream steps undetected until an inspection is performed. 3