BOAM: A Business Oriented Identification Approach of Microservices Within Legacy Systems

The microservices architecture (MSA) is highly popular for its scalability, deployability in the Cloud and compatibility with DevOps practices. Many companies are migrating their legacy systems to an MSA. They need to rely on automatic approaches to ease their migration while taking into account their business features. Existing migration approaches to an MSA often focus on technical features but neglect functional ones, which are essential for appropriate MS granularity. To address this lack, we introduce BOAM (Business Oriented identification Approach of Microservices), a hybrid approach that focuses on business decomposition by leveraging not only technical features, such as source code, but also business oriented artifacts, especially use cases. BOAM thus leverages static and semantic analyses of source code using nanoentities (data, operations or artifacts), followed by a semantic analysis of use cases to capture business features. For that, BOAM leans on machine learning, particularly clustering methods, to identify microservices through technical (source code) and business (use cases) artifacts. The goal is to ensure that identified microservices are technically sound and meet specific business features of the company. Our evaluation shows that BOAM outperforms other literature approaches to identify microservices, achieving an average precision of 74.51% and recall of 77.93%.