AI-Driven Practical Mainframe Modernization: A Safe, Scalable Path from Mainframe to Cloud

Google Cloud presented a strategy that modernizes mainframes in phases using AI and cloud, preserving functionality, rather than moving them to the cloud all at once in a 'big bang' approach. The key point is not simple COBOL-to-Java conversion, but transforming application dependencies, data models, interfaces, and data stores together.

Large-scale mainframe environments intertwine various elements, including legacy data stores such as DB2 and VSAM, transaction monitors like CICS and IMS TM, complex sequential workflows, protocols based on CTG, IMS Connect, MQ, and LU 6.2 Sockets, and dedicated operations utilities. Therefore, the modernization process must use actual production traffic to verify the functional equivalence of new code and reduce risk.

Google Cloud's approach consists of four pillars: assessment, modernization, de-risking, and data migration. It combines the code understanding and reasoning capabilities of Gemini models with dedicated mainframe modernization products to analyze and transform large-scale legacy environments.

The starting point, the Mainframe Assessment Tool (MAT), reverse-engineers the legacy codebase to provide the foundational information needed for modernization. Its key features are as follows.

• Dependency visualization across applications and data stores, including DB2 and VSAM
• Automated business rule extraction (BRE) that converts complex logic into requirements and decision trees
• Automated documentation that generates up-to-date technical documentation from production source code
• Domain and business function discovery that identifies application boundaries, business domains, and inputs, outputs, interfaces, and processing locations

MAT organizes the verified logic requirements of existing applications and business processes, enabling teams to design a cloud-native environment and drive incremental modernization.