Feedback-Directed Cross-Layer Optimization of Cloud-Based Functional Actor Applications

Even though actor-based frameworks such as Akka have been used by thousands of companies for over a decade, it is very difficult to configure cloud-based functional actor applications in a way to minimize their response time and to utilize provisioned resources with the maximal efficacy at the computing distributed cloud cluster nodes where these applications are deployed. As a result, the performance-tuning process is ad-hoc, laborious, manual, tedious and very expensive.We developed a novel low-code approach for cross-layer optimization of actor-based distributed applications. In our method, runtime information is automatically collected and analyzed in parallel to the executing application using search-based genetic algorithms. These algorithms generate changes to the application design, which are automatically applied to regenerate the application while it continues to service clients’ requests. We applied our approach to automatically created distributed Akka applications demonstrating its effectiveness in finding an optimal configuration.