Teradata, Oracle, SQL Server
Replacing a Teradata, Oracle or SQL Server warehouse with Databricks swaps capacity sized for peak demand for elastic, serverless compute. The hard part is converting procedural logic such as BTEQ, PL/SQL and T-SQL, which we do with LakeBridge, accelerators and engineer review.
Warehouse schemas, stored procedures, BTEQ and PL/SQL logic, reporting extracts
Each component has a clear target on the Databricks Data Intelligence Platform.
| Legacy warehouses | Databricks |
|---|---|
| Teradata, Oracle or SQL Server warehouse | Databricks SQL warehouses on Delta tables |
| BTEQ and TPT scripts | SQL and Lakeflow Declarative Pipelines |
| PL/SQL and T-SQL procedures | SQL and Python |
| Change data capture tools | Lakeflow Connect |
| Data marts | Gold layer tables and metric views |
Inventory every table, job, report and user. Map dependencies, find what is unused, and estimate effort and cost.
Output: Migration assessment and business caseDesign the target lakehouse, governance and security. Group workloads into waves by value and risk.
Output: Target architecture and wave planConvert SQL and ETL with LakeBridge and our accelerators, then review and test the output.
Output: Converted, tested code per waveRun old and new side by side and compare results table by table until they match.
Output: Signed-off reconciliation reportsMove users in waves, retire the old system, then tune cost and performance.
Output: Decommission plan and FinOps baselineWe use Databricks LakeBridge for profiling, code conversion and reconciliation, alongside our own accelerators and engineer review of every critical path.
Capacity-bound warehouses cost the most at peak. Elastic compute on the lakehouse changes the economics.
Read the article: Teradata, Oracle and SQL Server: replacing the legacy warehouse →Often, because you pay for compute when it runs instead of for peak capacity. We model the cost on your real workloads during the assessment.
We agree matching rules with business owners and reconcile old and new results table by table before any user switches.
