AWS re:Invent 2025 - Using generative AI to accelerate database modernization with AWS DMS (DAT420)

Accelerating Database Modernization with AWS DMS and Generative AI

Overview

• This presentation covers new advancements in AWS Database Migration Service (DMS) and the integration of generative AI to streamline database schema conversion.
• The speaker, Mike Rvit, is a database migration specialist who now works on the product team, focusing on how schema conversion is integrated with AI.
• Key announcements include:
  • Integration of generative AI into the AWS DMS schema conversion process
  • Demonstration of the Transform service for end-to-end .NET and SQL Server to PostgreSQL modernization

Database Schema Conversion Challenges

• Traditional schema conversion tools often provide misleading complexity metrics based on percentage of objects converted.
• Percentage-based metrics do not accurately reflect the true effort required, as a small percentage of a large codebase can still be a significant undertaking.
• Common issues like dynamic SQL can cause schema conversion tools to mark objects as difficult to convert, even when the actual conversion is straightforward.
• The presentation emphasizes the importance of diving deeper into the details rather than relying solely on percentage-based reports.

AWS DMS Schema Conversion with Generative AI

• The AWS DMS schema conversion tool now integrates generative AI capabilities powered by the Bedrock engine.
• Supported source databases include Db2, SQL Server, and Oracle, with PostgreSQL as the target.
• The conversion process involves two steps:
  1. Deterministic rules-based conversion using the established DMS engine
  2. Probabilistic conversion using the generative AI model if the deterministic rules fail

Demonstration: Cybase Database Modernization

• The presentation showcases the modernization of a Cybase database, a common legacy system in the finance industry.
• Schema conversion using the deterministic rules fails to convert a view and a stored procedure, resulting in error codes.
• The generative AI-powered conversion is then demonstrated, successfully converting the problematic objects.
  • The converted code is saved as a SQL script, allowing for manual review and testing before deployment.
  • The generated code includes informational comments to indicate the sections that were converted using the AI model.

Demonstration: End-to-End .NET and SQL Server to PostgreSQL Modernization

• The Transform service, announced by G2 at the conference, is demonstrated for a full-stack modernization from .NET and SQL Server to PostgreSQL.
• The process involves several steps:
  1. Analyzing the source .NET application and SQL Server database for complexity and dependencies
  2. Generating a migration plan with multiple waves based on complexity and interoperability
  3. Automating the conversion of the database schema and data using DMS
  4. Modifying the .NET application code to work with the new PostgreSQL database

Key Takeaways

• Generative AI can significantly enhance the capabilities of traditional schema conversion tools, handling complex transformations that would otherwise require manual effort.
• The integration of AI into the AWS DMS schema conversion process provides a more robust and reliable solution for database modernization.
• The Transform service demonstrates the potential for end-to-end modernization, automating the conversion of both the database and the application code.
• Careful testing and validation are still essential, as the AI-generated code may not always be perfect and may require manual review and adjustment.

Technical Details

• Supported source databases: Db2, SQL Server, Oracle
• Target database: PostgreSQL
• Deterministic rules-based conversion engine in DMS, in use for over 15 years
• Probabilistic conversion using the Bedrock generative AI model
• Transform service for full-stack .NET and SQL Server to PostgreSQL modernization

Business Impact

• Enables organizations to modernize legacy database systems, often critical to core business operations, with reduced effort and risk.
• Facilitates the migration of finance and other industry-specific applications that rely on legacy database technologies like Cybase.
• Provides a path for organizations to adopt modern cloud-based database platforms like PostgreSQL while preserving existing application investments.
• Accelerates the overall database modernization process, allowing organizations to realize the benefits of cloud-based databases more quickly.

Examples

• Demonstration of Cybase database modernization, including the use of generative AI to convert a problematic view and stored procedure.
• Walkthrough of the Transform service for end-to-end .NET and SQL Server to PostgreSQL modernization, including the automated conversion of the database schema and application code.