Comparing Software Testing Strategies Using Reliability Growth

Yinong Chen
Department of Computer Science
University of the Witwatersrand
Johannesburg, SOUTH AFRICA
yinong@cs.wits.ac.za

Jean Arlat
LAAS-CNRS
7 Avenue du Colonel Roche
31077 Toulouse, FRANCE
arlat@laas.fr

Abstract

This paper proposes an input domain-based reliability growth model.
Partition and random testing are used to generate inputs from ICD
(Input Case Domain). Input generation is generally considered to be
imperfect, i.e., it cannot generate inputs uniformly from the entire
ICD and therefore their fault coverage is reduced. Fault detection
and correction are also assumed to be imperfect, that is, the
probability to detect and fix a fault decreases with increasing
numbers of test runs. This copes with the fact that the remaining
faults after a period of testing are more difficult to detect and
to fix than the earlier faults. Based on the input domain-based
reliability growth model, the testing efficiency of partition and
random testing are compared.