Examination of Decomposition Techniques from Legacy Monolith to Softwarized Microservice-Oriented Architectures in Industry 5.0 Vision

Abstract

Purpose: In this study, monolith analysis methods, microservice identification, and decomposition methods proposed for the transition to microservice architectures that enable the development of appropriate solutions by adapting to the complex demands that will shape the technological infrastructure of the future are evaluated. Need for the study: Decomposition from monolithic architectures to microservices has become a popular approach in organizations and companies with Industry 5.0. This transformation of Industry 5.0 enables businesses to gain a competitive advantage and can provide a quick solution to personalized problems such as personal service systems. Methodology: The study, decomposition from monolith to microservice, initially includes monolith analysis, followed by microservice decomposition review. Various classification methods have been proposed for microservice identification and decomposition and are aligned with Industry 5.0 principles, focusing on artificial intelligence (AI)-based approaches, especially human-centered AI. Findings: Four analysis methods (domain, static, dynamic, and version) are identified for monolith analysis, with static and dynamic being the most common. Version analysis is not typically used alone. In the decomposition phase, clustering-based methods are prevalent due to the uncertain dimensions of microservices. Rule-based and unsupervised methods are identified for decomposition, with AI algorithms like affinity propagation, Kmeans clustering, hierarchical clustering, Hungarian algorithm, genetic algorithm, latent Dirichlet allocation (LDA), and minimum spanning tree (MST) being employed. Practical implications: Microservice architecture enables flexibility, scalability, and resilience compared to monolithic structures. Decomposing large-scale monolith projects into microservices is challenging, requiring selection of appropriate monolith analysis methods based on project details (e.g., domain analysis for detailed Unified Modelling Language (UML) diagrams) before proceeding with decomposition. This transformation improves deployment, maintenance, fault isolation, and scalability, while allowing for diverse service-specific databases and programming languages. © 2025 by Isil Karabey Aksakalli.