The Scenario: 50TB Oracle EDW, 500 Procs, 200 Reports, $50M Line
Estimated read: 3 min
Here’s the system you’re walking into for this capstone. A 50TB Oracle EDW, accumulated over roughly twenty years, running 500 stored procedures – the same number the opening section of this part used to make the point that a migration’s real scope is procedural logic, not just table row counts. Two hundred downstream reports read from it, ranging from a daily ops dashboard nobody would notice going stale for a day to a monthly board deck. And running through the middle of it is one order-to-cash pipeline that reconciles to a $50M-per-quarter revenue line in the general ledger – the one workload on the inventory where “close enough” was never an acceptable verdict, and won’t be here either.
You’ve done the work every prior section walked through. The workload inventory classified all 500 procedures and their dependent tables as Lift, Redesign, or Retire. The 3-R decision and the TCO calculator settled on a phased re-platform: straightforward CRUD procedures rehosted with minimal change, the handful of procedures with genuine business logic re-architected as Lakeflow Declarative Pipelines, and a documented list of Retire candidates that never made the trip. Schema translation ran through Lakebridge with the blind spots it doesn’t catch checked by hand. The reconciliation stack – count, sum, checksum, hash, and semantic parity – has been running nightly against every migrated table pair since day zero, not bolted on the week before cutover. The parallel run has been live for the contractual window, and the rollback procedure has been rehearsed against a deep clone, not just described in a document.
That’s not a coincidence – it’s the point of this capstone. Every artifact this part built exists so that today isn’t the first time anyone has looked at these numbers.
The $50M line is the detail that changes how every other decision in this scenario gets made. A migration with no revenue-critical workload can tolerate a longer rollback window, a slower go/no-go process, more tolerance for an open question at cutover. This one can’t – every lecture in this section assumes that constraint is real, not hypothetical, and every judgment call gets made against it.
What’s still ahead is the part no artifact can fully rehearse in advance: the live decision. The next lecture is the go/no-go decision matrix you walk into the war room holding – built in The Migration Playbook, and this is where it gets used for real, not as a template. After that, the war room itself: a live variance shows up during the reconciliation window, and it has to get diagnosed and resolved inside a fixed clock, with the business asking for a yes-or-no answer. The scenario is synthetic, but the shape of the decision – reconciliation status, rollback readiness, sign-off, compute headroom, one clock running down – is the same shape you’ll face on a real engagement, just compressed into a few lectures instead of a few months.
| ← Previous: The War Room: Capstone Simulation | Next: The Go or No-Go Decision Matrix → | |:—|—:|