Lune

ASE2025

Microservices Identification Using LLM

Jay Gandhi, Raveendra Kumar Medicherla, Manasi Patwardhan, Dipesh Sharma, Ravindra Naik

DOIPublisher

Abstract

  • Identification of microservices within legacy monolithic applications is a critical, challenging, expert-driven task. Existing techniques often cluster technical elements of the application without aligning them with business domain. These technically inclined approaches may not fully address the larger legacy modernization objectives. To address this issue, we propose a novel approach that leverages Large Language Model (LLM) to infer the domain intent of key technical elements within the architecture. Our approach then combines these intents with architectural dependencies, and clusters them using Graph Neural Network (GNN) to identify candidate microservices that are domain-coherent. Preliminary evaluation across four benchmark applications of varying sizes and domains demonstrates the promising potential of our approach.