AI-Driven Dependency Analysis for Migrating Monolithic Applications to Microservices Architecture

Abstract

Microservices Architecture (MSA) has become a de-facto standard for designing cloud-native enterprise applications due to its efficient infrastructure setup, service availability, elastic scalability, dependability, and enhanced security. Transitioning existing monolithic systems to microservices is essential to leverage these benefits. However, manual decomposition of large-scale applications is labor-intensive and prone to errors. AI-based systems offer promising solutions for automating this process. This paper introduces CARGO (Context-sensitive lAbel pRopaGatiOn), a novel un-/semi-supervised partition refinement technique that utilizes a context- and flow-sensitive system dependency graph of monolithic applications. CARGO refines and enhances the partitioning quality of existing microservice partitioning algorithms. Experiments demonstrate that CARGO improves partition quality, reduces distributed transactions, and enhances performance metrics such as latency and throughput in microservice applications. ​