Microservices Identification in Monolith Systems: Functionality Redesign Complexity and Evaluation of Similarity Measures

As monolithic applications grow in size and complexity, they tend to show symptoms of monolithic hell, such as scalability and maintainability problems. To help suppressing these problems, the microservices architectural style is applied. However, identifying the services within the monolith is not an easy task, and current research approaches emphasize different qualities of microservices. In this paper we present an approach for the automatic identification of microservices, which minimizes the cost of the monolith's functionalities redesign. The decompositions are generated based on similarity measures between the persistent domain entities of the monolith. An extensive analysis of the decompositions generated for 121 monolith systems is done. As result of the analysis we conclude that there is not a similarity measure, neither a combination of similarity measures, that provides better decomposition results in terms of complexity associated with the functionalities migration. However, we prove that it is possible to follow an incremental migration process of monoliths. Additionally, we conclude that there is a positive correlation between coupling and complexity, and that it is not possible to conclude on the existence of a correlation between cohesion and complexity.