Using Graph Convolutional Networks for Identifying Potential Microservice Candidates from Monolith Application
Meng-Sin Yang, Kuo-Hsun Hsu · 2024
Today, many companies are migrating from monolithic architecture applications to microservices architecture to enhance maintainability and scalability. Despite various proposed methods for decomposing microservices from monolithic applications, significant challenges persist, particularly in metrics such as granularity, and cluster size. Recent research has explored the use of graph neural networks (GNNs) to capture structural and behavioral relationships within software systems through graphical representations. In this research, we proposed incorporating Graph Convolutional Networks (GCN) to improve microservices decomposition, leveraging the benefits of graphical representation for a more precise granularity and effective clustering of functionalities into microservices. Our approach is evaluated through a comprehensive analysis of microservices decomposition to assess its feasibility and potential benefits.