Robustness Evaluation of Software Systems through Fault Injection

Abstract

Over the last decades, software has been introduced in desperate safety domains, such as automotive, avionics and railways, just to name a few. For these domains, software is demanded to be highly robust to hardware faults and software faults since its failure may endanger human life, harm the environment, or cause economical loss. Fault injection, the deliberate inoculation of faults, is a powerful means to assess the robustness of software components that goes far beyond traditional testing techniques. Fault Injection encompasses several techniques, among them robustness testing and software implemented fault injection, which emulate software faults and hardware faults, respectively. Despite the intensive use of these techniques, their application is still costly. This thesis focuses on robustness testing and software implemented fault injection, for both analyzes the �effect of workload on the experiment outcomes. Furthermore, the thesis suggests approaches to make the fault injection techniques more cost-effective by leveraging on the workload.