Use of failure-intensity models in the software-validation phase for telecommunications

Telephone switching systems require the use of more complex and bulky software. It is therefore important for France Telecom to take an interest in software quality before the operational phase. Trend tests show the evolution of detected failures during the validation phase. When these tests show a tendency towards improvement, we can apply software-reliability growth models. This paper describes experiments conducted at CNET, Lannion A Center, concerning the use of failure-intensity models in the software-validation phase of 3 telecommunication products. The experiments show that the logarithmic model is the most accurate. This paper also addresses future perspectives in data collection and use of reliability models.