Measuring software reliability
Shari Lawrence Pfleeger · IEEE Spectrum · 1992
The component concepts of the term 'reliability' are examined to clarify why its measurement is difficult. Two approaches to measuring the reliability of finished code are described. The first, a developer-based view, focuses on software faults; if the developer has grounds for believing that the system is relatively fault-free, then the system is assumed to be reliable. The second, a user-based view more in keeping with the standard IEEE/ANSI definition, emphasizes the functions of the system and how often they fail. The advantages of the failure-based approach are discussed, and various techniques are described. The issues that require further exploration and definition before reliability measurement becomes a straightforward for software as for hardware are identified.>