Databricks Migration Architect: Legacy EDW to Lakehouse

Descripción

One in three enterprise data warehouse migrations stall or get rolled back — not at extract-transform-load, but at trust. The tables arrive on Databricks and the business never moves, because the team treated a migration as data movement instead of logic, validation, and governance. This course is the playbook that stops that. You will own the exact path used to assess, convert, validate, and cut over a 50TB-plus enterprise EDW from Oracle, Teradata, or SQL Server to the Databricks lakehouse. Lakebridge automates the mechanical 80 percent of schema and SQL translation for free. This course teaches the 20 percent it cannot do — the part where the architect earns their salary. You will classify every workload as Rehost, Re-platform, Re-architect, or Federate with a defensible 3-year TCO model that survives CFO scrutiny. You will run Lakebridge schema and PL/SQL translation, then audit its blind spots — precision loss, transaction semantics, recursive macros, and window-function drift. You will decompose 30 years of stored-procedure business logic, architect ingestion with CDC, Auto Loader, and Lakeflow Declarative Pipelines under enforced data contracts, and prove semantic parity with a 5-layer reconciliation engine. Then comes the moment that decides the program: a zero-downtime, parallel-run cutover governed by a Go / No-Go decision matrix. You will rebuild legacy security in Unity Catalog, mapping 500 roles to 12 ABAC tags with masking and cross-engine access, and engineer the FinOps layer — compute arbitrage, system-table chargeback, and Predictive Optimization — so the migrated platform stays affordable. The course closes with a full War Room capstone simulation: sequence the whole migration end to end and defend every decision. This is an advanced, architect-level course. By the end you will not just move a warehouse — you will decide, validate, and defend the migration, which is the difference between a project that ships and one that lands in the migration graveyard.

Lo que aprenderás

• Assess any legacy EDW and classify every workload as Rehost, Re-platform, Re-architect, or Federate with a defensible 3-year TCO model.
• Convert schema and PL/SQL with Lakebridge, then audit its output for the precision loss, transaction semantics, and window-function errors it misses.
• Architect ingestion with CDC, Auto Loader, and Lakeflow Declarative Pipelines plus enforced data contracts so Bronze never becomes a swamp.
• Prove semantic parity with a 5-layer reconciliation engine and run a zero-downtime, parallel-run cutover using a Go or No-Go decision matrix.
• Rebuild legacy governance, mapping 500 roles to 12 ABAC tags, with masking and cross-engine access in Unity Catalog.
• Engineer the FinOps layer with compute arbitrage, system-table chargeback, and Predictive Optimization to keep the migrated platform affordable.

Requisitos

• Strong SQL — you can read and reason about complex stored procedures, joins, window functions, and aggregations.
• Some hands-on exposure to Spark or Databricks (notebooks, clusters, Delta tables); deep expertise is not required.
• A background working with a legacy enterprise data warehouse (Oracle, Teradata, or SQL Server) will help you get the most from the migration scenarios.