The single most common planning mistake is working backward from a renewal date with no margin. If your VMware contract expires in December, a migration that starts in September will fail, or you'll end up paying a bridge renewal at exactly the moment you have zero leverage. Map the phases below against your real calendar before you commit to a date.
The six phases
Discovery & assessment · 2–4 weeks
You can't migrate what you can't see. This phase builds the source-of-truth inventory and surfaces the surprises early, while they're cheap.
- Full VM inventory: vCPU, RAM, storage, OS, owner
- Application dependency mapping
- Licensing audit and renewal-date confirmation
- RTO/RPO and compliance requirements
- Platform shortlisting (2–3 candidates)
- Budget band established (cost guide)
Design & platform selection · 2–6 weeks
Decisions get locked here: target platform, architecture, network design, and the wave plan. Shorter for managed-VMware targets, longer when hardware must be specced and ordered.
- Final platform decision (compare options)
- Target architecture, sizing, and network/IP plan
- Storage and backup/DR design
- Migration tooling selection (HCX, Move, etc.)
- Wave groupings and go/no-go criteria
- Hardware ordered (lead times: 4–12 weeks)
Pilot · 2–4 weeks
Move a representative, low-risk slice, typically 5–10% of VMs, and validate the whole chain: tooling, runbook, performance, backup, rollback. The pilot exists to be wrong cheaply.
- Build target environment and connectivity
- Migrate dev/test and one tolerant prod workload
- Validate performance against baselines
- Test backup, restore, and rollback end to end
- Time the runbook, this calibrates the wave schedule
- Go/no-go review with stakeholders
Production waves · 4–16 weeks
The long middle. VMs move in dependency-grouped waves, usually 2–6 of them, lowest risk first, with validation gates between waves. This phase scales with VM count and downtime windows, not effort.
- Wave 1: remaining non-prod and tolerant workloads
- Middle waves: standard production by app group
- Final wave: crown jewels (ERP, databases)
- App-owner validation gate after every wave
- Rollback plan kept live until each gate passes
- Status comms at every milestone
Cutover & stabilization · 1–3 weeks
Final DNS/IP cutovers, the last replication sync, and a hypercare window while the environment settles. Most "migration failures" are actually skipped stabilization.
- Final cutover runbook executed (checklist)
- Performance baselines compared pre/post
- Backup jobs live on the new platform
- DR failover tested end to end
- Hypercare: daily checks, fast-path escalation
Decommission · 1–2 weeks (after a 2–4 week hold)
Only after every validation gate has passed: power down the old cluster, terminate licensing, and reclaim the budget line. Keep a final cold backup of the source environment regardless.
- Source hosts powered down after agreed hold period
- VMware subscription/SnS termination notice sent
- Hardware redeployed, sold, or recycled
- Documentation and runbooks finalized
- Lessons-learned review and TCO actuals captured
Typical total duration by environment size
| Environment | Replatform (Nutanix/Proxmox/Hyper-V) | Managed VMware / hyperscaler VMware | OpenShift / containerization |
|---|---|---|---|
| 25–100 VMs | 3–5 months | 6–12 weeks | 6–12 months |
| 100–500 VMs | 4–9 months | 2–5 months | 9–18 months |
| 500–2,000 VMs | 9–18 months | 4–9 months | 18+ months |
These are typical ranges for reasonably run projects with engaged app owners. VMware-to-VMware moves run shortest because VMs are replicated, not converted.
Where schedules actually slip
- App-owner availability. Validation gates wait on people, not technology. Book the sign-off calendar in phase 1.
- Hardware lead times. HCI nodes and switches can take 4–12 weeks; order during design, not after.
- The undiscovered dependency. The fax server nobody documented that talks to the ERP. Dependency mapping in discovery is what catches it.
- Change freezes. Retail Q4, fiscal year-end, audit windows, block them out before building the wave schedule.
- Renewal collision. Start at least 6–9 months before your VMware expiry so you're never forced into a panic renewal. If you're inside that window now, a managed provider bridge or short-term renewal may be part of the plan.
Want this mapped to your actual renewal date and VM count? A free assessment produces a dated plan, and the calculator pairs it with cost in minutes.