Distributed expectation-driven residual testing
Leila Naslavsky · 2004
Software testing techniques are widely used to assure properties such as dependability, as well as to gather system usage information to support software evolution. However, in spite of such techniques, comprehensive testing of commercial-grade software is rarely performed, resulting in software that is released with neglected test obligations. The dissemination power of the Internet and the increased connectivity of systems have made the pressure to release software earlier even more apparent. Fortunately, at the same lime, researchers have investigated using the infrastructure of the Internet for performing remote testing and analysis. One particular approach called residual testing allows the testing of software after it is deployed to beta testers or end users. This paper builds upon existing residual testing techniques, presenting an approach to integrate residual testing with a way to dynamically measure the behavioral conformity of the deplored system with its specifications. Our approach uses the facility of an extensible and configurable publish/subscribe infrastructure, dynamic instrumentation techniques and software tomography to address many of the issues faced by current remote residual testing strategies.