26 Oct 2026

Databricks Migration Services: Scope, Tools, and Process

Q: What are Databricks migration services?

They are services that move data, code, pipelines, security, and reports from legacy platforms onto the Databricks Lakehouse, covering assessment, a pilot, conversion, validation, cutover, and enablement.

Q: What are the 7 types of cloud migration?

The 7 Rs are retire, retain, rehost, relocate, repurchase, replatform, and refactor. A Databricks migration usually combines replatforming with selective refactoring.

Databricks migration services move data, code, pipelines, security, and reports from legacy warehouses, Hadoop, ETL tools, or other clouds onto the Databricks Lakehouse. A typical engagement covers assessment, a production pilot, data and code conversion, validation, phased cutover, and team enablement. They come from Databricks Professional Services, certified Brickbuilder partners, and specialist firms, and Databricks' free Lakebridge tool now automates much of assessment, SQL and ETL conversion, and reconciliation. What services add is judgment: deciding what to migrate, reproducing source-platform behavior, and proving the numbers match before cutover.

Searching for Databricks migration services returns three very different things: Databricks' own programs, certified partners, and specialist firms with their own tooling. This page explains what each offers, what the free Lakebridge tool already covers, and what you should still expect from a migration partner. If you are earlier in the decision, start with our overview of cloud migration to Databricks.

Quick Summary

Databricks migration services follow one arc: assess, pilot, convert, validate, cut over, and enable. Free tooling now handles much of the syntax conversion, so the value of a services partner lies in rationalization, source-platform semantics, security design, and proof that outputs match. Choose a partner by asking for a pilot on your own workload, not by comparing headline percentages.

Key Takeaways

Three provider types exist. Databricks Professional Services, certified Brickbuilder partners, and specialist firms, and many programs combine them.
Lakebridge is free. It covers an Analyzer, a Converter, and Reconcile, and Databricks states it can automate up to 80% of migration tasks, a vendor figure rather than a guarantee.
Tools convert syntax; engineers protect semantics. NULL handling, rounding, case sensitivity, and duplicate rows are where silent drift starts.
A pilot de-risks the program. Test the converters and the reconciliation method on one real workload before scaling.
Treat headline percentages with care. Claims of effort savings or zero downtime are vendor-reported, so ask how they were measured.
Naming has changed. Delta Live Tables is now Lakeflow Spark Declarative Pipelines, and Workflows is now Lakeflow Jobs.

Who Provides Databricks Migration Services

Provider Type	What They Offer	Best For
Databricks Professional Services	Migration guidance, full migration delivery, or migration assurance alongside system integrator partners, according to Databricks	Organizations that want the platform vendor involved from the start
Certified Brickbuilder migration partners	System integrators that use automated tooling and code converters to re-platform data, pipelines, and reports	Large, multi-workstream programs that need delivery scale
Specialist migration firms	Deep expertise in specific source platforms, with their own converters, test automation, and validation frameworks	Estates with hard source semantics, such as Teradata, Oracle, or Informatica, where exact output matching matters

DataTerrain sits in the third group, with 17+ years of ETL and data migration work. The categories overlap, and many organizations combine Databricks' own guidance with a specialist for conversion and validation.

What Databricks Migration Services Include

A full engagement follows the same arc whether Databricks, a partner, or a specialist delivers it:

Phase	What Happens	What You Should Receive
Discovery and assessment	Audit sources and map dependencies; inventory code, jobs, schedules, and downstream consumers, with usage data	A scoped inventory and a retire-versus-migrate list
Rationalization and architecture	Retire unused objects; design the lakehouse, including medallion layers, Unity Catalog security, cost controls, and CI/CD	A target architecture and a wave plan
Production pilot	Migrate one focused, end-to-end use case to test converters, data integrity, and performance	Evidence on automation rate and accuracy before scaling
Data and code modernization	Land data as Delta Lake; convert SQL, procedures, and ETL into Spark SQL, PySpark, or Lakeflow pipelines	Converted, tested code and migrated data
Validation and phased cutover	Reconcile outputs against the source; run parallel or staged cutovers with consumers moved in the same wave	A signed-off reconciliation and a controlled switch
Enablement and optimization	Train teams, tune performance and cost, retire legacy systems, and optionally add managed services	A trained team, cost guardrails, and retired licenses

Databricks' own guidance describes a hybrid of lift-and-shift and modernization as the most common successful path, so the strategy decision belongs in the first phase, as covered in our Databricks migration overview and key checklist for BI modernization. For the reporting layer on top, see how BI automation simplifies report migration and our report conversion services.

Lakebridge: What the Free Tool Covers, and What Still Needs Engineering

Lakebridge, announced by Databricks in June 2025, is a free, open-source migration toolkit from Databricks Labs. It integrates technology from BladeBridge, a migration engine Databricks acquired earlier in 2025. Databricks states that it can automate up to 80% of migration tasks and cut project timelines roughly in half. Treat those as the vendor's figures for favorable cases, not a promise for your estate.

Component	What It Does	What Still Needs People
Analyzer	Profiles the source and produces an inventory of objects to migrate	Deciding what to retire, sequencing waves, and sizing effort
Converter	Translates legacy SQL and ETL code into Databricks SQL or PySpark using plugins (Morpheus, BladeBridge, and the experimental Switch); an agentic converter using Genie Code is in beta	Behavior that syntax cannot capture: NULLs, rounding, ordering, and procedural logic that must be redesigned
Reconcile	Compares source and target data to prove the move	Choosing what to compare, resolving every difference, and signing off before cutover

Databricks lists support for more than 10 source systems, including Teradata, Oracle, SQL Server, Netezza, Snowflake, and Azure Synapse. Coverage changes between releases, so check the current Lakebridge documentation for your dialect and version. Our legacy systems to Databricks article lists these tools alongside the surrounding migration approach.

The practical question is therefore not whether to use Lakebridge. It is who owns everything around it: the retire decisions, the architecture, the security redesign, and the proof that results match.

Where Silent Data Discrepancies Hide

The expensive migration failures are not crashes. They are reports that run and return slightly different numbers. These are the usual causes:

Source Behavior	What Changes on Databricks	Safeguard
Row-by-row procedural code (PL/SQL, T-SQL, Teradata BTEQ) using cursors and temp tables	Literal translation runs slowly and expensively on Spark	Refactor into set-based Spark SQL, PySpark, and Delta MERGE operations, as in our PL/SQL conversion article
NUMBER, DECIMAL, DATE, and TIMESTAMP handling	Precision loss and rounding at exact midpoints drift totals	Enforce type-mapping standards and compare at column level against real data
Empty strings (Oracle treats an empty string as NULL)	Spark treats an empty string and NULL as different values, so filters, joins, and counts change	Normalize explicitly and test for it
Case sensitivity and duplicates (Teradata NOT CASESPECIFIC columns and SET tables)	Delta tables compare case-sensitively and keep duplicate rows by default	Reproduce the old behavior deliberately in keys, joins, and dedupe logic
Indexes, distribution keys, partitioning, and optimizer hints	None of these apply to Spark, so key queries can miss their SLAs	Use liquid clustering, data skipping, and Photon, and tune against real workloads
Hidden consumers (reports, extracts, and applications)	Anything reading the legacy platform breaks at cutover	Inventory every consumer and move it in the same wave

This is where tooling and services divide. A converter can translate the syntax of a cursor loop, but only comparison against real data shows whether the result matches, the same principle behind automated ETL testing with Python and the difference between traditional and modern ETL.

Migration Paths by Source

Source	Typical Databricks Target	Read Next
Teradata, Oracle, SQL Server, and Netezza warehouses	Databricks SQL on Delta Lake with medallion layers	Data warehouse to Databricks
Hadoop, Hive, and other legacy estates	Delta Lake with Unity Catalog on cloud storage	Legacy systems to Databricks
Informatica, DataStage, SSIS, and Talend ETL	Lakeflow Spark Declarative Pipelines (formerly Delta Live Tables), PySpark, and Lakeflow Jobs	ETL migration to Databricks
Alteryx workflows	PySpark or Databricks notebooks	Alteryx vs Databricks
Snowflake, Redshift, Synapse, and other clouds	Lakehouse Federation to query in place, then migrate workload by workload	Cloud migration to Databricks
Oracle PL/SQL procedures	Set-based Spark SQL and PySpark	Legacy PL/SQL to modern ETL

For the whole-estate view, see our data platform migration to Databricks article, and for how Databricks compares with other targets, our Databricks vs Snowflake comparison. One naming note: Databricks renamed Delta Live Tables to Lakeflow Spark Declarative Pipelines and Workflows to Lakeflow Jobs, and existing pipeline code keeps working without changes.

Choosing AWS, Azure, or Google Cloud

Databricks runs on all three major clouds, so the cloud decision usually comes before the Databricks decision. Choose based on where your data, identity, and security tooling already live, then use that cloud's native transfer services for bulk and incremental loads. If AWS is your target, our AWS ETL services and mainframe-to-AWS migration articles cover the surrounding data movement, and our data lake work covers the storage layer.

How to Choose a Databricks Migration Partner

Ask for a pilot on your own workload. A pilot shows the real automation rate and accuracy, which a sales deck cannot.
Ask what is automated and what is manual. Every provider automates the easy majority, so the useful answer is how the hard remainder is handled.
Ask how outputs are reconciled. Look for column-level comparison against the source on real data, not a visual check.
Check source-platform depth, not just Databricks depth. Silent drift comes from the legacy platform's behavior, so the team must know it well, as we do in our Informatica consulting work.
Check governance and cost design. Unity Catalog security, cluster policies, auto-termination, and tagging should be designed before the first workload moves.
Treat headline numbers with care. Claims of percentage savings or zero downtime are vendor-reported, so ask how they were measured and on what kind of estate.

For a broader view of what works in automation, see our ETL automation for legacy data migration article and our Informatica PowerCenter vs IICS comparison for teams planning ETL moves alongside this one.

A Related Spark Migration: Alteryx to PySpark on AWS

This engagement ran on AWS EMR Serverless rather than Databricks. It is included because DataTerrain applies the same Spark engineering and validation approach to Databricks migrations.

A global manufacturer of electronic connectors and interconnect solutions engaged DataTerrain to migrate its Alteryx workflow portfolio to PySpark on AWS.

Business challenges

A large portfolio of complex workflows, including iterative macros and hierarchical data expansion
Large data volumes with multi-source joins and conditional logic
Significant recurring licensing costs
Zero tolerance for data discrepancies in production output

What DataTerrain delivered

Complete migration of all workflows to PySpark on distributed cloud infrastructure
Metadata-level analysis of every workflow before conversion
Custom solutions for rounding, NULL handling, sort tiebreakers, and Unicode stripping
Column-by-column validation across every output table
Parallel production runs confirming data accuracy before cutover

Results

Migration delivered at a fraction of the estimated effort
Licensing costs eliminated
Every output column validated, with zero unresolved data discrepancies
The customer continues to expand Spark adoption in the cloud

Cost and Timeline Factors

Cost and duration depend on four things: historical data volume, the number of code objects (procedures, scripts, and ETL jobs), the number of reports and applications to repoint, and how strict the validation must be. An automated inventory gives a far better estimate than counting tables. A pilot typically runs in weeks, and enterprise programs are usually delivered in waves over several months. Pricing models vary, so confirm scope and rates directly with each provider.

Frequently Asked Questions

What are Databricks migration services?

They move data, code, pipelines, security, and reports from legacy platforms to the Databricks Lakehouse, covering assessment, a pilot, conversion, validation, cutover, and enablement.

What is Databricks Lakebridge, and is it free?

Lakebridge is Databricks' free, open-source migration toolkit with an Analyzer, a Converter, and Reconcile. Databricks states it can automate up to 80% of migration tasks.

What are the 7 types of cloud migration?

The 7 Rs: retire, retain, rehost, relocate, repurchase, replatform, and refactor. A Databricks migration usually combines replatforming with selective refactoring.

What are the top data migration tools for Databricks?

It depends on the job: Lakebridge for assessment, conversion, and reconciliation; Lakehouse Federation for querying data in place; Auto Loader and Lakeflow Connect for ingestion; and each cloud's native transfer services for bulk loads.

Why are people leaving Databricks?

There is no reliable public data on customers leaving, and the question mostly reflects forum discussion. Teams that reconsider typically weigh cost, complexity, and fit with their existing cloud stack, so test your own workloads before deciding.

Prove the Numbers Match with DataTerrain

Lakebridge converts the syntax. The hard part is proving the numbers still match when your business moves to Databricks. DataTerrain brings 17+ years of experience and 400+ customers in data and reporting modernization, plus AWS partnership, to Databricks migrations covering assessment, conversion, security design, validation, and cutover, with flexible hours and no long-term binding contracts.

Talk to our Databricks migration team

Contents