Migration Mental Model
1. Migration has multiple dimensions
A data-platform migration changes more than storage:
- Data — records, history, keys, volumes, retention.
- Semantics — measures, dimensions, currencies, units, hierarchies.
- Logic — transformations, derivations, business rules.
- Integration — sources, deltas, interfaces, schedules.
- Security — identities, roles, row-level restrictions, sharing.
- Consumption — SAC, APIs, applications, extracts.
- Operations — monitoring, recovery, ownership, SLAs.
- Economics — capacity, licensing, runtime, support effort.
A migration is complete only when the required business capability works on the target and the old dependency is demonstrably removable.
2. Modernization patterns
| Pattern | Meaning | Typical use |
|---|---|---|
| Lift | Move with minimal logical change | Reduce infrastructure ownership |
| Shift | Move workload to a target service with limited redesign | Accelerate cloud adoption |
| Refactor | Redesign models and pipelines | Remove technical debt |
| Rebuild | Create a new target model | Major semantic or domain redesign |
| Coexist | Run source and target together | Incremental migration |
| Retire | Remove obsolete workload | Cost and complexity reduction |
Do not label a project “migration” until the target outcome and the retirement boundary are explicit.
3. Architect's first question
Ask: What must remain identical, and what is allowed to improve?
Financial reconciliation may require identical numbers while the physical model, pipeline, security design, and consumption layer can change.
4. Migration factory
Use a repeatable flow:
Discover → Assess → Classify → Design → Pilot → Migrate → Reconcile → Cut over → Stabilize → Retire
Every wave should produce evidence, not just a completion ticket.