Method-Level Syntactic and Semantic Clustering for Microservice Discovery in Legacy Enterprise Systems

Enterprise systems, such as enterprise resource planning, customer relationship management, and supply chain management systems, are widely used in corporate sectors and are notorious for being large, inflexible and monolithic. Their many application-specific methods are challenging to decouple manually because they manage asynchronous, user-driven business processes and business objects having complex structural relationships. We present an automated technique for identifying parts of enterprise systems that can run separately as fine-grained microservices in flexible and scalable Cloud systems. Our remodularization technique uses both semantic properties of enterprise systems, i.e., domain-level business object and method relationships, together with syntactic features of the methods’ code, e.g., their call patterns and structural similarity. Semantically, business objects derived from databases form the basis for prospective clustering of those methods that act on them as modules, while on a syntactic level, structural and interaction details between the methods themselves provide further insights into module dependencies for grouping, based on K-means clustering and optimization. Our technique was prototyped and validated using two open-source enterprise customer relationship management systems, SugarCRM3 and ChurchCRM. The empirical results demonstrate improved feasibility of remodularizing enterprise systems, inclusive of coded business objects and methods, compared to microservices constructed using class-level decoupling of business objects only. Furthermore, the microservices recommended, integrated with “backend” enterprise systems, demonstrate improvements in execution efficiency, scalability, and availability.