Getting Started with Cloud Migration
The Six Rs of Migration
Every workload has a migration strategy. Pick deliberately:
- Rehost ("lift and shift") โ fastest, least change, preserves technical debt
- Replatform โ minor changes (managed DB, container runtime) for medium gain
- Repurchase โ replace with SaaS where viable
- Refactor โ rebuild for cloud-native; biggest payoff, biggest risk
- Retire โ delete what nobody uses
- Retain โ keep on-prem if cloud doesn't earn its keep
A Workable Sequence
Phase 1 โ Discovery and Inventory
Catalog every application, dependency, data flow, and SLA. Most teams discover 20โ30% more than they expected.
Phase 2 โ Pilot
Pick a non-critical, well-understood workload. Migrate it end to end. Document the gaps. Calibrate your estimates.
Phase 3 โ Foundation
Build the landing zone before mass migration:
- Multi-account/project structure with IAM baselines
- Network topology (VPCs, peering, transit gateways)
- Logging and monitoring from day one
- CI/CD pipelines and secret management
- FinOps tagging and budget alerts
Phase 4 โ Wave Migration
Group workloads into waves by dependency clusters. Migrate a wave, validate it, move on. Avoid frozen change windows by running cloud and on-prem in parallel where needed.
Phase 5 โ Optimisation
Once everything is running, the real work starts:
- Right-size compute and storage
- Replace lifted services with managed equivalents where it pays
- Establish FinOps reviews and chargeback
What Goes Wrong
- Migrating data without a strategy. Egress, change-data-capture, cutover windows โ plan upfront.
- Underestimating networking. Hybrid networking is harder than the slides suggest.
- Skipping observability. Without logs and metrics, you can't tell if "it works" or "it works for now."
- Treating it as one project. Migration is a programme; expect 12โ24 months for a real estate of any size.
The Bottom Line
Migration succeeds when it's treated as a portfolio of decisions, not a single big-bang move. Pilot, learn, wave, optimise.
*We've migrated workloads from on-prem to AWS, Azure, and GCP. Get a migration plan โ*