"Leveraging Generative AI for Mainframe Application Modernization to Cloud-Based Architectures"

"Mainframe platforms continue to support missioncriticalenterprise workloads; however, growing requirementsfor scalability, agility, and cost efficiency are accelerating theadoption of cloud-native architectures. Modernization of legacyCOBOL-based systems remains challenging due to monolithicprogram structures, complex interdependencies, and limiteddocumentation. This paper proposes a Generative Artificial Intelligence(GenAI)-driven modernization framework that systematicallytransforms simple and complex COBOL, COBOL\/DB2,and batch-oriented mainframe applications into cloud-readymicroservices. The framework integrates automated applicationassessment, semantic code understanding, business rule extraction,AI-assisted refactoring, and cloud-native service generation.Experimental evaluation using representative enterprise batchapplications demonstrates significant reductions in manual effortand modernization cycle time while preserving business semantics.The proposed approach enables incremental, scalable, andlow-risk migration of legacy systems to cloud-based architectures."

大型主机平台仍在支撑关键企业业务工作负载；然而，企业对可扩展性、敏捷性与成本效益的需求日益增长，正加速云原生架构的采用步伐。基于COBOL语言的遗留系统现代化改造仍面临诸多挑战，其根源在于单体式程序结构、复杂的相互依赖关系以及文档的匮乏。本文提出了一种生成式人工智能（Generative Artificial Intelligence, GenAI）驱动的现代化改造框架，可将简单与复杂的COBOL、COBOL/DB2以及面向批处理的大型主机应用系统性地转换为适配云环境的微服务。该框架集成了自动化应用评估、语义代码理解、业务规则提取、AI辅助重构以及云原生服务生成等功能。通过使用典型企业批处理应用开展的实验评估表明，该方法可在保留业务语义的前提下，大幅减少人工工作量并缩短现代化改造周期。所提出的方案可实现遗留系统向云架构的增量式、可扩展且低风险的迁移。