Reliability-growth analysis for an Ada-coding process

In a recent software development effort, an ARINC Research Corporation development team translated 500 system transactions into 11000 lines of code in Ada-compilable program design language (PDL). The team established rigorous data collection and review processes to collect and classify defect data and identify reliability growth in the PDL being produced. Two classes of defects for the coding process were defined: minor defects and logic defects. Data were collected by transaction, by defect type, and by lines of PDL. The defect analysis indicated that the learning process, and thus reliability growth, were different for the two defect classes. In addition, a period of reduced work load and changes in coding standards affected the observed defect rates for each class differently. As a result of this analysis effort, a database was established to track the coding process and to provide a basis for estimating various parameters of future efforts (e.g., time and cost) as well as the impact of process changes