Automatically and transparently distributing the execution of regression test suites

From the perspective of the software testing practitioner, regression testing is often an important technique for developing high quality software systems. The software testing research community has developed several different approaches in order to ensure that regression testing process is cost-effective and practical. The techniques developed by Harrold, Rothermel, Rosenblum, Weyuker, Soffa and others have attempted to increase the cost-effectiveness of regression testing by intelligently selecting a subset of an entire regression test suite for execution. Recent approaches created by Elbaum, Malishevsky, Rothermel, Untch, Chu, and Harrold have tried to improve a regression test suite’s rate of fault detection by prioritizing the execution of the tests. In this paper, we describe the conceptual foundation, design, and implementation of an approach that distributes the execution of regression test suites. We show that our technique can complement existing regression test selection and prioritization approaches. We also demonstrate that it can be used independently when existing approaches are not likely to increase regression testing efficiency. Finally, we describe the design and implementation of Joshua, an application that facilitates the distributed execution of regression test suites for Java-based software systems.