Microservice decomposition and evaluation using dependency graph and silhouette coefficient

• Ana Santos PUC Minas
• Hugo Paula PUC Minas

Abstract

The benefits provided by microservices architecture in some application scenarios are a motivating factor for organizations to migrate their monoliths to this architecture. Extracting microservices from existing monolithic code bases presents a key challenge in this context, and there is a lack of tools that automate not only the decomposition processes but also the evaluation of the resulting architecture. This work presents a new approach for microservice decomposition that analyzes source code of a monolithic application and, with the combined use of approaches in the literature, suggests parts to be extracted in microservices considering the artifacts: classes, methods and/or history of modifications. The quality of the microservices’ suggestions are assessed, quantitatively, through the silhouette coefficient, a quality metric used in clustering analysis, and the microservice granularity. A tool was developed to automate the process of microservice decomposition for Java repositories. As a result, it was observed that the tool generated clusters with satisfactory results and can be used as an auxiliary instrument by experts during the migration process from monolithic architecture to microservices.

Keywords: microservices, decomposition, monolithic application