Legacy Migration to Databricks

A field guide for migrating a legacy enterprise data warehouse (Oracle, Teradata, SQL Server, PL/SQL) to the Databricks Lakehouse: assessment, TCO, schema and procedure translation, reconciliation, cutover, governance migration, and FinOps.

In Part 1 you learned the Databricks primitives – Delta Lake, Unity Catalog, the medallion architecture, Lakeflow. In Part 2 you used them to build StepRight end to end on a clean slate. Part 3 is the job you’ll actually be hired for: there is no clean slate, only a twenty-year-old Oracle or Teradata warehouse with five hundred stored procedures, a nightly batch window that stopped fitting in its window three years ago, and a CFO asking why the license renewal costs more than the team that maintains it. This part walks the full arc of an enterprise EDW migration in the order a real engagement runs it: profiling the legacy estate and picking the right migration strategy for each workload, translating schema and procedural logic with Lakebridge and knowing exactly where it needs a human, standing up ingestion and CDC pipelines with Lakeflow, proving semantic parity with a reconciliation stack before anyone trusts a cutover, migrating governance into Unity Catalog, and running the FinOps discipline that keeps the new platform cheaper than the one it replaced. It closes with a capstone war-room simulation where you make the go/no-go call yourself.

Sections

#	Section	Items
1	Why Enterprise Migrations Fail (And the Architect Who Stops It)	5
2	The Autopsy: Profiling the Legacy Monolith	6
3	The 3-R Decision and the TCO That Convinces the CFO	6
4	Lakehouse Federation: Migrate Without Migrating	5
5	Schema Translation with Lakebridge (and Its Blind Spots)	6
6	Physical Design for Delta: Liquid Clustering Over Indexes	6
7	The Procedure Autopsy: Decomposing PL/SQL	6
8	Pattern Translation: Cursors, Triggers, Temp Tables, MERGE	7
9	AI-Assisted Migration: Lakebridge plus LLMs With a Human Gate	7
10	The Ingestion Decision Tree	6
11	CDC and Lakeflow Declarative Pipelines With Data Contracts	7
12	The Reconciliation Stack: Proving Semantic Parity	6
13	Building the Reconciliation Engine	6
14	The Parallel Run and Zero-Downtime Cutover	7
15	Unity Catalog and the Privilege Migration	6
16	ABAC, Masking, and Cross-Engine Governance	6
17	The Cost Iceberg and Compute Arbitrage	6
18	FinOps Engineering: System Tables and Chargeback	7
19	The Migration Playbook: Sequencing the Whole Thing	6
20	The War Room: Capstone Simulation	6

---

| ← Previous: Final Stage - Production Deployment and Smoke Testing Your Project | Next: Why Enterprise Migrations Fail (And the Architect Who Stops It) → | |:—|—:|

---

Table of contents

• Why Enterprise Migrations Fail (And the Architect Who Stops It)
• The Autopsy: Profiling the Legacy Monolith
• The 3-R Decision and the TCO That Convinces the CFO
• Lakehouse Federation: Migrate Without Migrating
• Schema Translation with Lakebridge (and Its Blind Spots)
• Physical Design for Delta: Liquid Clustering Over Indexes
• The Procedure Autopsy: Decomposing PL/SQL
• Pattern Translation: Cursors, Triggers, Temp Tables, MERGE
• AI-Assisted Migration: Lakebridge plus LLMs With a Human Gate
• The Ingestion Decision Tree
• CDC and Lakeflow Declarative Pipelines With Data Contracts
• The Reconciliation Stack: Proving Semantic Parity
• Building the Reconciliation Engine
• The Parallel Run and Zero-Downtime Cutover
• Unity Catalog and the Privilege Migration
• ABAC, Masking, and Cross-Engine Governance
• The Cost Iceberg and Compute Arbitrage
• FinOps Engineering: System Tables and Chargeback
• The Migration Playbook: Sequencing the Whole Thing
• The War Room: Capstone Simulation