Leveraging Graph-Based for Enhanced Service Identification and Microservices Partitioning in Business Process Systems
Kurnia Cahya Febryanto, Riyanarto Sarno, Abdullah Faqih Septiyanto, Kelly Rossa Sungkono, Shoffi Izza Sabilla, Sholiq Sholiq · 2024
Business process systems face critical challenges in identifying and partitioning microservices, particularly in Enterprise Resource Planning (ERP) systems where complex interdependencies and dynamic workflows make it difficult to establish optimal service boundaries. Traditional service identification approaches often rely on manual analysis or basic clustering techniques, lacking the capability to effectively capture and analyze intricate service relationships and dependencies. Furthermore, existing methods struggle to provide quantifiable metrics for assessing service quality and optimizing service boundaries in complex business processes. To address these limitations, this research introduces a novel graph-based approach utilizing Neo4j for enhanced service identification and microservices partitioning. The approach incorporates three key innovations: (1) automated extraction of service dependencies from Business Process Model and Notation diagrams to reduce manual effort and potential errors, (2) comprehensive structural analysis through graph database capabilities for deeper understanding of service relationships, and (3) advanced quality metrics evaluation framework combining multiple dimensions including cohesion, coupling, granularity, and maintainability. Experimental evaluation across multiple ERP modules demonstrates the effectiveness of the approach, showing significant improvements in service autonomy (up to 94 percent), maintainability (80–85 percent enhancement), and modularity (up to 100 percent improvement). While achieving optimal service boundaries and improved system comprehension, the approach reveals an important trade-off between system maintainability and service cohesion that provides valuable insights for future research directions in microservices architecture optimization.