Home → Engineering & Data Projects → Provincial Cutover ETL
Provincial Cutover ETL (6M Users)
This case study covers a province-wide telecom cutover ETL serving roughly 6 million users. By flattening deeply nested Oracle SQL, using hash joins, pre-filtering large historical tables, and orchestrating steps via shell into a rollback-safe DAG, the overall cutover window was reduced from about 4 hours to roughly 2 hours.
Dashboard
Data refresh: On-demand (historical replay)
What you’re seeing: a compact view of key cutover steps, record counts, and elapsed time along the ETL path — from source tables to target structures. The goal is to answer one question quickly: “Can we migrate millions of users inside a safe window, with a tested rollback if anything goes wrong?”
Share on X
Share on LinkedIn
Share via Email
How to read & interact
- Follow the ETL stages: move through source extraction, staging, transformation, and final load to see how many records flow through each step.
- Watch the timing: compare elapsed time per stage to identify where de-nesting SQL and hash joins delivered the biggest runtime savings.
- Check safety margins: review how much buffer remains inside the cutover window and how rollback checkpoints are distributed across the DAG.