Why Migration Fails (and How to Avoid It)
Most cloud migrations fail not because of technical complexity, but because of inadequate discovery and planning. I've seen organisations rush to "lift and shift" only to discover their on-premise architecture simply doesn't map to GCP — and they end up paying more than they did on-premise.
The checklist below is what I walk through with every business client before we write a single line of Terraform.
Phase 1 — Discovery
- Inventory all workloads: VMs, databases, storage buckets, message queues, batch jobs.
- Map dependencies: Which services talk to each other? What are the latency requirements?
- Classify data: PII, regulated data, and compliance requirements (GDPR, HIPAA, etc.).
- Assess network topology: Internet-facing vs internal. Existing VPN or interconnect needs.
Phase 2 — Strategy Selection
Choose the right "R" for each workload:
- Rehost (Lift & Shift): Move VMs as-is using Migrate to VMs. Fast, low risk, low gain.
- Replatform: Move to managed services (e.g., Cloud SQL instead of self-managed MySQL).
- Re-architect: Redesign for cloud-native — GKE, Cloud Run, Pub/Sub. Highest gain, most effort.
Phase 3 — GCP Environment Setup
- Establish a Resource Hierarchy: Organisation → Folders → Projects
- Configure IAM least-privilege: No broad Editor roles in production.
- Set up VPC networks, subnets, and firewall rules before migrating any workload.
- Enable Cloud Logging, Cloud Monitoring, and Security Command Center.
- Define billing alerts from day one.
Phase 4 — Migrate & Validate
Always migrate in waves — start with non-critical workloads to build confidence and surface surprises. Run parallel environments for at least one billing cycle before cutting over production.
Phase 5 — Post-Migration Optimisation
- Right-size compute using Recommender API.
- Apply Committed Use Discounts (CUDs) once baseline usage is stable.
- Review network egress costs — often the biggest surprise bill item.
Want me to walk through your specific workloads? Book a free consultation.
Engineer Resilient Multi-Cloud AI Systems
We architect autonomous AI agents, RAG pipelines, and cloud infrastructure across GCP and AWS.
Request a Technical Scope