Empirical assessment of architecture-based reliability of open-source software

A number of analytical models have been proposed earlier for quantifying software reliability. Some of these models estimate the failure behavior of the software using black-box testing, which treats the software as a monolithic whole. With the evolution of component based software development, the necessity to use white-box testing increased. A few architecture-based reliability models, which use white-box approach, were proposed earlier and they have been validated using several small case studies and proved to be correct. However, there is a dearth of large-scale empirical data used for reliability analysis. This thesis enriches the empirical knowledge in software reliability engineering. We use a real, large-scale case study, GCC compiler, for our experiments. To the best of out knowledge, this is the most comprehensive case study ever used for software reliability analysis. The software is instrumented with a profiler, to extract the execution profiles of the test cases. The execution profiles form the basis for building the operational profile of the system, which describes the software usage. The test case failures are traced back to the faults in the source code to analyze the failure behavior of the components. These results are used to estimate the reliability of the software, as well as the uncertainty in the reliability analysis using entropy.

Name thesis:degree_name

MS

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Lane Department of Computer Science and Electrical Engineering

Year dc:date.available

2004

Author dc:creator

• Perugupalli, Ranganath

Contributors dc:contributor

• Katerina Goseva-Popstojanova.

• Computer science

Identifier

https://researchrepository.wvu.edu/etd/1555

OAI identifier oai:identifier

oai:researchrepository.wvu.edu:etd-2558