Skip to content
TechFabric

Migration accelerator

Migrate the estate, and be able to prove it

A governed migration factory for moving warehouse and ETL estates to Databricks.

3 common questions, answered below ↓
Role in the family
Migration accelerator
Databricks surfaces
LakebridgeUnity CatalogDatabricks Asset BundlesDatabricks AppsTemporal
Status
Console, worker, CLI, and declarative Databricks Asset Bundle deployment are implemented. Twenty-seven source platforms have executable playbooks, including Snowflake, Azure Synapse, Amazon Redshift, Oracle, Teradata, SQL Server, BigQuery, Netezza and SAP. The first two weeks of an Airlift engagement are sold on their own as the Migration Readiness Sprint.

The problem

Conversion is the easy part of a migration. The hard part is proving what was in scope, which tool version produced which artifact, what evidence says it is correct, and how to reverse a cutover at two in the morning when it is not.

How it works

Airlift composes Databricks Lakebridge for profiling, analysis, SQL conversion, and reconciliation, then wraps it in a policy-enforced lifecycle: discovery and scope acceptance, independent validation, signed migration certificates, approved waves, and reversible cutover. Every artifact carries the tool version that produced it and the evidence that cleared it.

What it does

Evidence, not assertions

Each converted artifact is bound to immutable evidence proving it is ready to promote.

Signed migration certificates

A wave cuts over only when its certificate is issued. The certificate is the gate, not a status field.

Reversible cutover

Checkpoint, apply, verify, with a defined path back out of an external change.

Wave planning

Scope is accepted explicitly and cut over in approved waves.

What it changes for you

A warehouse migration your risk function will actually sign off. Scope, evidence, certificates, and a reversible cutover are part of the process, so the migration stops being a long weekend with a rollback plan nobody has tested.

Where this shows up in an engagementMigrations to Databricks

The conversion works. The evidence does not exist.

Converting SQL is the part everybody plans for. What sinks a migration is the six months afterwards, when somebody asks a question the project cannot answer.

Nobody can say what was in scope

The inventory was a spreadsheet, taken in March, by somebody who has left. Three objects nobody counted are still running on the old platform and something depends on them.

The cutover has no way back

Which is why it happens on a Friday night. A plan with no reverse is a plan that can only be run when being wrong is survivable, and that window is always a weekend.

The old platform never gets turned off

Nobody is funded for decommissioning, so the bill runs in parallel for a year and the migration that was supposed to save money has not yet.

None of that is a conversion problem. It is a record-keeping problem, and a migration that keeps its own records answers all three without anybody being asked to remember.

The shift

Three ways to move a warehouse. Two of them lose the receipts.

Every migration picks one of these, and the choice is usually made by whoever is available.

Hand conversion

Accurate, and it does not finish

Inventory by hand
Convert object by object
Spot-check
Big-bang weekend
Hope

A conversion tool and a spreadsheet

Fast, and unprovable

Tool converts
Spreadsheet tracks
Versions drift
Cutover
Reconcile for months

A governed migration factory

Provable, and reversible

Inventory as data
Executable playbook
Signed certificate per object
Validated in waves
Reversible cutover

What runs the migration

Six parts, and the reason it is a factory rather than a project is that each one produces a record.

Twenty-seven source platforms

Snowflake, Synapse, Teradata, SQL Server, Oracle, Hadoop, Redshift and twenty more, each with an executable playbook rather than a methodology slide.

Inventory as data

Object counts, dependencies and the pieces nobody has run in two years, held as a queryable record rather than as a spreadsheet somebody owns.

Signed migration certificates

Per object, recording what was converted, by which tool version, and what the validation said. An auditor reads them instead of asking you to remember.

Validation before the switch

Row counts, checksums and business-rule reconciliation between the old estate and the new one, signed off before any traffic moves.

A reversible cutover

The switch has a tested way back. That is what turns a migration weekend into a migration morning.

Decommissioning in the plan

Turning the old platform off is where migrations quietly fail, because nobody is funded for it. It is a phase here rather than an afterthought.

27

Source platforms with executable playbooks

Built because we were doing it by hand

Airlift exists because we ran these migrations before we automated them, and the parts that hurt are the parts it now handles. It is free and open source under Apache-2.0 and it runs in your own workspace.

FAQ

TechFabric Airlift, answered

Which source warehouses can Airlift migrate?

Twenty-seven, each with an executable playbook. Snowflake, Azure Synapse and Teradata have their own pages at /databricks/from-snowflake, /databricks/from-synapse and /databricks/from-teradata. Redshift, Oracle, SQL Server, BigQuery, Netezza and SAP are covered too. Tell us your source and we will confirm what the playbook does today.

What makes a migration reversible?

Cutover runs as checkpoint, apply, verify, with a defined path back out of an external change. A wave only cuts over once its certificate is issued, so the gate is the evidence itself.

Can our risk function audit the migration?

That is the design goal. Every converted artifact is bound to immutable evidence and carries the tool version that produced it, so what was in scope, what proved it correct and who approved the wave are all answerable after the fact.