Performance exploration of the REST API using online GAN

Abstract

Performance Testing is a critically important step in the application development process. Manually achieving performance exploration comes with high human-cost and, in general, inefficiency. Software testing automation is a research area with many contributions which observes frequent advancements. In the context of the SBST 2022 CPS tool competition, Åbo Akademi University laboratory has published a new Machine Learning technique that merges a GAN-based model and the online learning process: Online Generative Adversarial Network (OGAN).

This thesis approached the performance exploration problem using the OGAN algorithm on particular applications: REST APIs. This thesis sets the required context and develops the information background concerning: REST APIs principles, the Performance Testing process and the usage of Deep Neural Network, specifically Generative Adversarial Networks (GANs), in an online learning approach. The Machine Learning technique tackled: OGAN is presented and its algorithm detailed and discussed. Finally, an application of this technique is proposed and explored on the PetClinic REST API.