City Research Online

Abstract

The standard approach to deriving the confidence bound for the probability of failure on demand (pfd) of a software-based system is to perform statistical tests on the whole system as a “black-box”. In practice, performing tests on the entire system may be infeasible for logistical reasons, such as lack of availability of all component subsystems at the same time during implementation. This paper presents a general method for deriving a confidence bound for the overall system from successful independent tests on individual system components. In addition, a strategy is presented for optimizing the number of tests allocated to system components for an arbitrary system architecture that minimizes the confidence bound for the system pfd. For some system architectures, we show that an optimum allocation of component tests is as effective as tests on the complete system for demonstrating a given confidence bound. The confidence bound calculation makes use of many of the concepts used in the reliability analysis of hardware structures, but unlike a conventional hardware analysis, the method does not presume statistical independence of failures between software components, so the confidence bound calculation for the software should always be conservative.

Publication Type:	Article
Additional Information:	© 2020. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/
Publisher Keywords:	Statistical testing, Confidence bounds, Software reliability, Fault tolerance
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Science & Technology > Computer Science > Software Reliability
SWORD Depositor:	Symplectic Administrator

Export

Downloads