Predicting Method Crashes

Software monitoring tools have high performance overheads because they typically monitor all processes of the running program. This overhead can be dramatically reduced by lowering the number of methods being monitored to a smaller subset. We present an approach that can help arrive at such a subset by reliably identifying methods that are likely to crash in the future. Our approach involves learning patterns from features of methods that previously crashed to classify new methods as crash-prone or crash-resistant. An evaluation of our approach on two large open source softwares, ASPECTJ and ECLIPSE, showed that we can correctly classify crashprone methods with an accuracy of up to 80%. Notably, we also found that the classification models can also be used for crossproject prediction with virtually no loss in classification accuracy. In a further experiment, we demonstrate how a monitoring tool, RECRASH could take advantage of only monitoring crash-prone methods and thereby, reduce its performance overhead and maintain its ability to perform its intended tasks.