Assessing reliability of modular software (Q1124324)

A Markovian model of a modular software system is presented. The software system has M modules, designed for execution of K tasks. Transitions between modules are assumed to follow a Markov process. Assuming known failure rates for modules, the probability of error free execution of a program, and overall failure rate are derived. With the objective of minimizing development/testing cost, a procedure is given for determining module failure rates that enable the system to a function at a prescribed failure tolerance. The procedure is illustrated through the use of an example exhibiting modular software with a single program. The ideas of potential use to project managers and software engineers during software development.