The microservice-oriented architecture (MSA) is an architectural style which involves organizing an application as of small independent services, each oriented towards one business functionality while being data autonomous. In pursuit of modernizing their software to take advantage of the Cloud, companies have been eager to migrate their monolithic legacy software towards an MSA. This migration necessitates an identification phase to reorganize classes around the monolith’s functionalities as a set of microservice candidates. However, most identification approaches fail to utilize the monolith’s internal multilayered architecture to identify those functionalities, and thus the microservices. As a consequence, ignoring the internal multilayered architecture increases the risk of identifying microservice by their technical layer which is recognized as a conceptual anti-pattern. In this paper, we explore the impact of the multi-layer architecture in monolithic applications during the identification to develop a semi-automatic approach that relies on it to identify an MSA. Particularly, we analyze the presentation layer to determine the endpoints of each business functionality of the monolith. From these endpoints, we apply a vertical decomposition to identify the necessary classes to implement each feature as a microservice. In the process, we also define the bounded context of each microservice during the vertical decomposition of the data-access layer. For the evaluation, we implemented a model-driven process and applied it on a set of varying open-source applications commonly used in the literature. We compared the results of approach with and without the reverse-engineering of the internal architecture to measure the impact of our approach on the identification of quality microservices. Using decomposition metrics (e.g., MoJoFM, c2ccvg), we were able to measure a significant positive impact.

International audience

Due to its tremendous advantages, microservice architectural style has become an essential element for the development of applications deployed on the cloud and for those adopting the DevOps practices. Nevertheless, while microservices can be used to develop new applications, there are monolithic ones, that are not well adapted neither to the cloud nor to DevOps. Migrating these applications towards microservices appears as a solution to adapt them to both. In this context, we propose an approach aiming to achieve this objective by focusing on the step of microservices identification. The proposed identification, in this paper, is based on an analysis of the relationships between source code elements, their relationships with the persistent data manipulated in this code and finally the knowledge, often partial, of the architect concerning the system to migrate. A function that measures the quality of a microservice based on its ability to provide consistent service and its interdependence with others microservice in the resulting architecture was defined. Moreover, the architect recommendations are used, when available, to guide the identification process. The conducted experiment shows the relevance of the obtained microservices by our approach.