Are Longer Test Sequences Always Better? - A Reliability Theoretical Analysis

One of the interesting questions currently discussed in software testing, both in practice and academia, is the role of test sequences on software testing, especially on fault detection. Previous work includes empirical research on rather small examples tested by relatively short test sequences. Belief is "the longer the better", i.e., the longer test sequences are, the more faults are detected. This paper extends those approaches applied to a large commercial application using test sequences of increasing length, which are generated and selected by graph-model-based techniques. Experiments applying many software reliability models of different categories deliver surprising results.