Before You Migrate: Ask the Right Question
Over the last decade, I've worked on multiple data modernization initiatives involving Oracle, SAP HANA, PostgreSQL, Vertica, traditional data warehouses, Hadoop ecosystems, and modern cloud platforms. One question consistently comes up:
"How do we migrate to a modern cloud platform without creating another expensive legacy platform in the cloud?"
Unfortunately, many organizations treat migration as a technology replacement exercise. It is not.
The wrong question:
"How do we move Oracle/HANA/Teradata to Snowflake?"
The right question:
"What should our target-state data platform look like in 3-5 years?"
Migration should never be driven by:
- End of support deadlines
- Vendor pressure
- Executive mandates alone
- Cloud-first checklists
Migration should be driven by:
- Scalability — can we 10x without re-architecture?
- Cost optimization — pay for what we use, not what we provision
- Faster analytics — from data to insight in minutes, not weeks
- Self-service reporting — empower business, don't gatekeep
- AI readiness — support RAG, ML, and agents from day one
- Data democratization — governed access, not data silos
Three Migration Approaches
Lift & Shift
Move schema and jobs with minimal changes. Fastest path but carries legacy design into cloud. Use only as a temporary landing zone.
Replatform
Keep business logic, modernize infrastructure. ETL→ELT, on-prem→Airflow, legacy BI→Power BI/Tableau. Sweet spot for most enterprises.
Modernization
Rebuild around business domains and data products. Future-proof, AI-ready, governed. Best long-term ROI. Requires architectural maturity.
Decision Matrix
| Factor | Lift & Shift | Replatform | Modernization |
|---|---|---|---|
| Cost | Low | Medium | High (upfront) |
| Time | Fast (weeks) | Medium (months) | Longer (quarters) |
| Risk | Medium | Medium | Higher (managed) |
| Long-Term Value | Low | Medium | Very High |
| AI Readiness | None | Partial | Full |
| Governance | Inherited (poor) | Improved | Designed-in |
| Cost Optimization | Marginal | Good | Excellent |
Key Areas Often Ignored During Migration
1. Data Quality
2. Metadata and Lineage
Most migration failures happen because nobody knows where data originated, who owns it, or which report consumes it. If lineage is not documented before migration — expect surprises post go-live.
3. Security and Access Model
Do not migrate permissions blindly. Redesign from scratch:
- RBAC — role-based, not user-based
- Row-Level Security — region, department, entity
- PII Protection — tokenization, dynamic masking
- Zero-trust — private endpoints, MFA, audit trails
4. Cost Management
Architecture Principles I Follow
Migration Do's & Don'ts
Do's
- Establish target-state architecture first — tech migration without architecture vision creates cloud-based legacy
- Classify workloads — not everything deserves migration. Retire what's unused
- Migrate business domains incrementally — HR, Finance, Sales, Marketing as separate waves
- Automate testing — validate record counts, data quality, business KPIs, and performance
- Involve business stakeholders — success is measured by business adoption, not tables migrated
Don'ts
- Don't migrate everything — enterprises typically discover 30-40% of assets are unused
- Don't migrate technical debt — cloud is not a garbage relocation service
- Don't ignore governance — design it before ingestion, not after
- Don't replicate old ETL patterns — modern architectures favor ELT
- Don't optimize only for today's reporting — build for AI workloads tomorrow
The Three Questions
Before migrating any object, ask:
- Does this data still provide business value?
- Would I design it this way if starting from scratch today?
- Will this support AI and analytics requirements over the next five years?
If the answer is "No" to any of these — redesign before migration.
Migration Readiness Scorecard
📋 Rate Your Organization (1-5 each)
- Data Quality Assessment Completed ___/5
- Data Lineage Available ___/5
- Business Ownership Defined ___/5
- Security Model Designed ___/5
- Cloud Cost Model Established ___/5
- Target Architecture Approved ___/5
- AI/ML Roadmap Identified ___/5
- Data Governance Framework Established ___/5
Final Thoughts
The most successful migrations I've been part of were never database migrations. They were platform transformations.
The organizations that realize the highest value use migration as an opportunity to:
- Modernize architecture
- Improve governance
- Simplify operations
- Prepare for an AI-driven future
The cloud platform is excellent. But success is rarely determined by the technology. Success is determined by the architecture decisions made before the first table is migrated.
And as architects, that is where our focus should always remain.