DEVCLASS AD
AWS has revamped its AI-based Transform tool, which assists with application modernization and migration, adding the ability to move from Microsoft’s SQL Server to PostgreSQL, VMWare to EC2 (Elastic Compute Cloud) and more, evidence of a new battleground as cloud giants use AI to move customers from competing products and platforms.
AWS Transform is the overall product name for a diverse set of tools. Transform for .Net was first previewed as a feature of Amazon Q Developer, aiming to migrate .Net Framework applications, which run only on Windows, to cross-platform .Net Core (now just called .Net). Microsoft also has tools for this purpose but AWS has long taken a particular interest in this type of migration since it enables customers to move away from Windows server with its license fees and friction with the largely Linux-based AWS platform.
The first release of AWS Transform for .Net was in May this year, but it did not attempt to migrate applications from using a SQL Server database, commonly used with .NET applications. The new release now does this, enabling AWS to call it “full stack Windows modernization.”
DEVCLASS AD
The new capability adds an option to migrate from SQL Server to Aurora PostgreSQL, which is technically described as “PostgreSQL-compatible” rather than being actual PostgreSQL instances as is the case with Amazon RDS (relational database service). The difference is that Aurora PostgreSQL uses distributed system techniques and has its own software and hardware optimizations which AWS claims gives up to three times the throughput of standard PostgreSQL. The downside is that it is more expensive in most cases, though Aurora can be more cost-effective when a high number of provisioned IOPS (input/output operations per second) are required.
The documentation reveals limitations. SQL Server must be running on AWS, not on premises or with a different hosting provider. SQL Server integration services, for data integration and transformation solutions, are not migrated. AWS Transform also classifies its tasks into categories: standard SQL has a high automation success; advanced T-SQL (the SQL Server query language) which requires human review and may need intervention; and features such as CLR (.Net) assemblies or complex full-text search that cannot be directly migrated and will need refactoring.
Along with the revamped Transform for .NET, AWS introduced new capabilities for other types of migration at its re:Invent conference running this week in Las Vegas. Transform for VMWare, which assists with migrating VMWare infrastructure and applications to native AWS services, has a new agentic AI discovery tool that can go on to automate new network configurations and server migrations. The same tool can also be used with other hypervisors such as Hyper-V, Nutanix and KVM.
Transform also has a new custom capability, where users can add their own transformation tasks via agentic AI. Unlike other Transform features, there is an additional charge for custom capabilities, based on per-minute fees for agent work.
DEVCLASS AD
AWS Transform for mainframe, which like the .Net variant was first released in May, has been updated with a new “reimagine” option, explained as rethinking the source application’s architecture, such as converting a monolithic design to one using microservices. There is also a new automated test plan generation, with both features driven by AI.
Many, probably most organizations struggle with legacy code that holds back other desirable changes, whether that is to do infrastructure, licensing costs, or application architecture. There is often a tension between dedicating time to refactoring this old code, and developing new features demanded by product teams. The notion of just running a migration tool is therefore compelling, but may well fail because of intricate dependencies that are hard to remove. Agentic AI is less constrained than purely algorithmic conversions, though it also introduces uncertainty because of the non-determinative nature of generative AI.
The benefit for the provider, when the tools work well, is the profit from increased use of the target products and services. Users should be wary of the tools over-provisioning, or selecting pricier options on the basis that they are technically superior, even though a cheaper option might be a better fit for the specific use case.
According to AWS cost consultant Corey Quinn, AWS even at one point inserted a clause into its service terms requiring that AWS Transform users “run the transformed workload on AWS for a minimum of 24 months,” a clause that was then hastily removed, but still a reminder of why these tools exist.
