An Input Domain-Based Reliability Growth Model and Its Applications in Comparing Software Testing Strategies
Yinong Chen* and Jean Arlat
LAAS-CNRS, 7 Avenue du Colonel Roche
31077 Toulouse Cedex - FRANCE
Full Report Paper in Postscript File
Abstract
Existing input domain-based reliability models do not account for
software reliability growth, because they do not consider fault corrections.
This paper proposes an input domain-based reliability growth model with fault
correction history being taken into account. Both partition and random testing
can be used to generate input cases for test runs. It is generally considered in
the model that input case generation, fault detection and fault correction can
all be imperfect. As an application of the proposed input domain-based
reliability growth model, the efficiency of random and partition testing in
terms of reliability growth is studied and compared analytically. The impacts on
the efficiency of testing strategies due to the number of faults in the program,
the distribution of fault in the input domain of the program, as well as the
imperfections of input case generation, fault detection and fault correction are
studied. Through sophisticated analysis we obtain some new results which explain
why and under which conditions random testing has the same efficiency as, a
higher or lower efficiency than partition testing. As a further application of
the proposed input domain-based reliability growth model, the efficiency of
variants of partition testing are studied and compared.
Keywords: software fault model, imperfect fault correction, software
reliability model, testing coverage, comparison between testing strategies
* On leave from the Department of Computer Science, University of the Witwatersrand, Johannesburg.
He is a postdoctoral fellow at LAAS-CNRS, Toulouse, under the European Commission Human Capital and Mobility Programme, CaberNet.