Automating the Recognition of Microservices from Business Process Analysis

Abstract

In the evolution from monolithic to modular software architectures, the integration of microservices is increasingly preferred due to its enhancement of operational efficiency and adaptability. Microservices are distinguished by their compact size, cohesive functionality, and independent operation, positioning them as a formidable alternative to traditional architectures. However, identifying optimal microservice configurations poses significant challenges. This study introduces an innovative automated method for deriving microservices from business processes, incorporating a broader BPMN with more activities and dependencies than previous studies to verify the quality of our approach. Our method leverages a cooperative clustering strategy that integrates multiple models to analyze control, information, and semantic connections within business operations. Each dependency type is addressed by a dedicated clustering algorithm, allowing modules to leverage insights from one another, thereby enhancing microservice identification accuracy. We validate our methodology through a case study in a bike rental service, involving 25 distinct activities to evaluate our dependency model’s effectiveness. The findings demonstrate superior precision in microservices generation compared to existing methods, highlighting the advantage of using detailed business operations as a foundation for precise microservice architecture design.