Migrating a Large Data Reporting Platform

How to choose an approach, and how to shape the project around it.

This is a living guide. It is revised and extended as I have more to say, rather than fixed at a point in time. Every page carries its own review date, and changes are listed in the changelog.

These are my own views, written from years of building and moving large Oracle data platforms. They are not Oracle’s official guidance and this is not Oracle documentation. Volatile product claims are quarantined in Product facts that will date so you can check them against current documentation rather than trusting a page I wrote once.

Start with who this is for if you want the scope, or pick a reading path below.

Reading paths

“I have been handed a migration and do not know where to start.” Copying the data is the easy part → Two kinds of warehouse → Do the discovery before you commit to the business case → Option zero

“I need to choose a method.” Complexity, not volume → Four options → My defaults → Model complexity dimensions

“I need to shape the project.” Waves or big bang → Nothing survives twelve hours → Making an architected platform decomposable → The consumption tail

“We are approaching cutover.” Rehearsal → Performance problems are the rule → Cutover: abort point and rollback boundary → Validation and handover

“It is live and the numbers do not match.” The numbers will not match