Evaluation of a software reliability improvement methodology

A methodology for improving software reliability based on the US Air Force's Independent Verification and Validation Program for critical nuclear control software is evaluated. The methodology is presented as a numerical simulation, whose evaluation can easily be seen to be non-dimensional, so that its conclusions are general. The analysis makes use of the "waterfall" model of software development, in which it is assumed that faults produced in the various phases are subsequently detected by using a suite of activities, techniques and tools (AT&Ts) with certain degrees of effectiveness in fault detection. By the end of the development, most faults are eliminated, leaving a residual which can produce system failures in operation. The percentage of faults remaining is dependent on the particular AT&Ts used and their intensity of use. Finally, the methodology yields a framework for assessing the costs of using AT&Ts vs. their effect on residual faults and, consequently, on operational reliability.