We show how software reliability predictions can increase confidence in the reliability of the NASA Space Shuttle Primary Avionics Software. These predictions, along with other methods of reliability assurance, such as inspections and fault tracking, provide a quantitative basis for achieving reliability objectives. Without a quantitative reliability assessment, software managers have no objective basis for deciding whether the software has been tested sufficiently to be ready for a flight or whether a serious failure during flight is likely. Our prediction methodology provides bounds on test time, remaining failures, program quality, and time to next failure that are necessary to meet Shuttle software reliability requirements. We also show that there is a pronounced asymptotic characteristic to the test time and program quality curves that indicate the possibility of big gains in reliability as testing continues; eventually the gains become marginal as testing continues. We conclude that the prediction methodology is feasible for the Shuttle and other safety critical applications.
展开▼
