Workload-Based Clustering of Coherent Feature Sets in Microservice Architectures

In a microservice architecture, each service is designed to be independent of other microservices. The size of a microservice, defined by the features it provides, directly impacts its performance and availability. However, none of the currently available approaches take this into account. This paper proposes an approach to improve the performance of a microservice architecture by workload-based feature clustering. Given a feature model, the current microservice architecture, and the workload, this approach recommends a deployment that improves the performance for the given workload using a genetic algorithm. We created Micado, an open-source tool, in which we implemented this approach, and applied it in a case study on an ERP system. For different workloads, the resulting generated microservice architectures show substantial improvements, which sets the potential of the approach.