Databricks Genie Migrate 正式 GA：支持 7 种数据仓库 SQL 转 ANSI SQL

Companies are racing to build AI agents and applications. Every one of these use cases depends on easy access to their data, but for many enterprises, the data AI needs is locked inside of legacy data warehouses. Migrations are traditionally expensive and time-consuming, so many companies stay put and miss out on using AI to its full potential.

Genie Migrate is changing that equation. Genie Migrate for Lakehouse is a specialized coding agent, built on Genie Code, that automates the end-to-end migration lifecycle from legacy data warehouses. We are excited to announce that Genie Migrate now supports converting SQL code from 7 data warehouses to open ANSI SQL, all generally available: Microsoft SQL Server, Azure Synapse, Teradata, Oracle, Snowflake, Amazon Redshift, and Google BigQuery.

Your entire migration in one place

Genie Migrate for Lakehouse takes on the repetitive work of a migration, like assessing what files to migrate first, converting SQL scripts, and determining why conversions are failing, so your engineers can focus on reviewing and approving results. Four features keep the experience simple, from the first upload to the last approval:

One place for the whole migration: Genie Migrate organizes your work in a migration project, a workspace that keeps every file, its status, and its converted output together, so the whole team can track progress in one place.
See how your code connects: Review lineage in your migration project to see which tables and files depend on each other, so your team knows what has to move together before you convert anything.
One click to convert: Click Convert and Genie Migrate converts files in parallel in the background.
Review in one view: Compare your original and converted code side by side in the SQL editor. Ask Genie Code in plain language to adjust anything that needs a judgment call, then approve files one at a time or in bulk.

Databricks migrations are already getting faster as teams bring AI into the work. The median migration time from onboarding to production at Databricks has dropped by 25% over the past 6 months. Customers can retire costly legacy platforms and stop paying for them much sooner than before with the help of Genie Migrate.

Accurate and effective migrations to Databricks Lakehouse

Genie Migrate converted more than 90% of files correctly on benchmarks for all seven sources. There are three factors that drive high accuracy:

Built-in knowledge of each source dialect: Genie Migrate is built with knowledge of each dialect's proprietary functions and procedural code. It uses this knowledge to convert code to their Databricks equivalents.
A check-and-revise loop: Genie Migrate automatically checks semantic correctness and syntax for its own output. When a check fails, it uses the error to diagnose the problem, revise the code, and check again.
Your team's custom standards: Custom skills allow you to customize Genie Migrate’s conversion output. If a conversion doesn't match your preferences, or the same issue keeps coming up, add a custom skill and Genie Migrate applies it to every file it converts.

The result is open ANSI SQL that runs on Databricks, follows your team's standards, and keeps the original logic, so you move from conversion to validation faster.

How MIXI is accelerating their BigQuery migration

While in Beta, Genie Migrate has already helped customers move to the lakehouse faster. MIXI is a Japanese digital entertainment, sports, and lifestyle company. Its MONSTER STRIKE team, which runs one of Japan's biggest mobile games, is moving the game's data mart pipelines from BigQuery to Databricks. The team runs approximately 120 daily BigQuery data marts covering user attributes, item masters, logins, gameplay, and purchases. For the migration, the team needed to keep its project conventions, such as how tables are named, how each daily job filters to its date, and how data types are converted.

To migrate to the Lakehouse, MIXI rolled out Genie Migrate across their entire data estate. The straightforward experience made it easy to get started and scale conversions across their data marts. With custom skills, the team could also easily apply project-specific standards, accelerating the conversion from GoogleSQL. The agent handled much of the repetitive work, from analyzing source code and applying the team’s rules to converting code, checking the results, and retrying when it found an issue. This kept the process simple and reduced manual effort for the team.

“Genie Migrate gave us a fast, effective path for moving pipelines from BigQuery to Databricks. 85% of the files we tested ran successfully without manual edits, shifting our time from labor-intensive translation to reviewing and refining results. By combining AI-powered conversion with our project-specific standards, we were able to migrate quickly, consistently, and at scale.” - Taiki Watanabe, Data Engineer, MONSTER STRIKE Operations Division at MIXI, Inc.

Genie Migrate’s ease of use was matched by the accuracy of the conversions. Of the converted files the team tested, approximately 85% ran without errors and required no manual edits. The remaining tested files needed only simple one-line fixes. Genie Migrate consistently applied the team’s instructions across files, producing accurate, reviewable code. This allowed the engineers to focus on validating results, enabling the two-person team to efficiently advance their data mart migration with confidence.

Get started with Genie Migrate

Genie Migrate for Lakehouse is generally available today for all seven data warehouse sources. Support for ETL sources is coming soon. The quickest way to see what Genie Migrate can do is to try it on your own code: create a migration project in your Databricks workspace, upload a few SQL files, and compare the converted code side by side.

When you're ready to plan the full migration, contact your Databricks account team or read the Genie Migrate for Lakehouse documentation for supported sources, prerequisites, limits, and the end-to-end workflow.

FAQ

What’s Genie Migrate for Lakehouse?

Genie Migrate for Lakehouse is an AI coding agent from Databricks, built on Genie Code, that automates data warehouse migrations to the Databricks Lakehouse. It runs inside your Databricks workspace. It maps dependencies between files and tables, converts proprietary SQL to open ANSI SQL in parallel, checks and fixes its own output, and lets engineers compare and approve every converted file. Genie Migrate for Lakehouse is generally available.

Which data warehouses can Genie Migrate convert to Databricks?

Genie Migrate for Lakehouse converts SQL from seven data warehouses to open ANSI SQL on Databricks: Microsoft SQL Server, Azure Synapse, Teradata, Oracle, Snowflake, Amazon Redshift, and Google BigQuery. All seven sources are generally available. Support for ETL sources is coming soon.

How accurate is Genie Migrate's SQL conversion?

On benchmarks, Genie Migrate converted more than 90% of files correctly for each of the seven supported sources. It checks its own output for syntax and semantic correctness, revises the code when a check fails, and applies your team's custom skills to every file it converts. At MIXI, about 85% of the tested BigQuery files ran without manual edits, and the rest needed only one-line fixes.

How long does a data warehouse migration to Databricks take?

The median Databricks migration reduces timelines by 25%, as teams bring AI into migration work. Genie Migrate shortens the code conversion phase by converting files in parallel in the background, so engineers spend their time reviewing and approving results instead of rewriting SQL by hand.

How do I migrate a legacy data warehouse to Databricks with Genie Migrate?

Genie Migrate runs a data warehouse migration in five steps inside your Databricks workspace:

1. Create a migration project.
2. Upload your source SQL files.
3. Review lineage to see which tables and files need to move together.
4. Click Convert. Genie Migrate converts the files in parallel in the background.
5. Compare the original and converted code side by side in the SQL editor, adjust anything with Genie Code or a custom skill, and approve files one at a time or in bulk.