Performance and Cyclomatic Complexity Correlation in Java Reactive Frameworks

Abstract: Software performance is of interest to all who want to lower their hardware costs and provide services with speedy responses to users, to this end reactive programming can be employed. Therefore it is important to measure the performance of tools such as reactive frameworks and to compare and contrast them with each other in order to improve the development of these tools, and help provide developers with the information they need when searching for a reactive framework that suits their project. To achieve this the study will aim to indicate the reproducibility of the research on reactive framework performance conducted by Ponge et al. on three commonly used reactive frameworks. Further, a root cause analysis to identify the sources of the identified bottlenecks will be carried out and complemented by suggestions for improving the performance of those parts in the reactive frameworks that are causing performance issues. An analysis of how performance correlates with the cyclomatic complexity of the frameworks will also be presented. We find, that in some test cases, the performance of the frameworks differs depending on the machine used, and that the latest framework versions do not show a marked performance increase. Further, we discover hotspots in the Mutiny framework and reason about potential synchronization bottlenecks. We attempt to find a novel use case for cyclomatic complexity as an indicator for performance but find that no correlation exists between cyclomatic complexity and performance for reactive frameworks.