ARCHITECTURE BASED SOFTWARE RELIABILITY PREDICTION FOR SAFETY CRITICAL SYSTEMS

Abstract

Software reliability is the measure of working ability of any software system without encountering failures. Predicting software reliability before the system is actually deployed in field can be propitious in many ways. This becomes even more vital in case of safety critical system where cost of failure includes loss of life. In traditional methods of determining the software reliability, we either wait till the system is deployed and then use its usage profile to determine reliability or develop a model of the system and perform simulation. These methods have limitations when applied to safety critical systems. To address those limitations, software reliability prediction techniques based on system architecture for component based software systems have been used lately. These methods basically include state based formal methods like Markov chains, path based or combinatorial models like fault trees and hybrid models. This report provides an overview of architecture based reliability prediction methodology, reviews the related literature to identify the research gaps. To address some of the research gaps, we have used a reliability prediction method based on Petri nets. The approach has been applied to predict the reliability of Shut Down System-1 used in nuclear power plant. To the best of our knowledge, no reliability prediction method based on Petri nets has been applied on the chosen system. We have compared the predicted reliability values with that obtained from system requirement specifications