Supported migration & DR paths
Not just VMware-to-Proxmox. Here's every source/destination pair Migration Center supports today.
| Source | Destination | Migration | Disaster recovery |
|---|---|---|---|
| VMware (vCenter/ESXi) | Proxmox | Yes | Yes - continuous sync, failover, test failover, failback |
| VMware (vCenter/ESXi) | VMware (another ESXi host) | Yes | Not yet - failover/failback aren't available for this pair |
| OpenStack Beta | Proxmox | Yes | Yes - continuous sync, failover, test failover (no failback to OpenStack) |
| OpenStack Beta | OpenStack (another endpoint) | Yes - first copy, continuous sync, cut-over | Not yet - failover and test failover aren't available for this pair |
| Proxmox | Proxmox (another node/cluster) | Yes | — |
| Hyper-V | Proxmox | Yes - first copy, repeat syncs and cut-over (a running VM is copied from a checkpoint); change tracking copies only the blocks that changed | Yes - continuous sync with change tracking, failover, test failover (no failback to Hyper-V) |
| Hyper-V | Hyper-V (another host) | — | Orchestrates Hyper-V's own native Replica: initial sync, test / planned / unplanned failover, reverse, health |
Beta marks paths that are built and covered by automated tests against simulated environments, and are still being proven on live infrastructure. Hyper-V (migration, DR replication and Replica orchestration) has been validated with testing.


Agentless block-level copy
VMware (vCenter or standalone ESXi) and OpenStack sources, moved to Proxmox - or to another VMware or OpenStack endpoint - without installing anything inside the guest. No agent to deploy, license, or clean up afterward.
- VMware over the vSphere API, to Proxmox or another ESXi host
- OpenStack over API and SSH, with Ceph, LVM, or NFS backend support, to Proxmox or another OpenStack endpoint
- Proxmox → Proxmox moves, for consolidating or rebalancing clusters you already run
Thin copy, changed blocks, compression & deduplication
Only the data that matters crosses the wire: after the first copy, changed-block tracking ships just what changed, and empty blocks are skipped where the destination is already zeroed. No extra SKU to buy.
- Repeat syncs and the final cut-over move only the changed blocks, for migration and DR alike
- Optional zstd wire compression per policy (off by default) - used when both ends have zstd, otherwise the copy simply runs uncompressed
- Optional deduplication per policy (off by default): on the first copy to Proxmox, a 1 MiB block already on the node - earlier in the same disk or from another VM in the policy - is copied locally on the node instead of being sent again. Every local copy is read back and checked, and anything doubtful is sent normally
- Live savings shown per job and on the performance dashboard: the thin-copy ratio, the compression ratio when it is on, and the deduplicated bytes when deduplication is on. With it off, the job still shows how much of the data repeated, so you can see whether it would pay off
Online, offline, or with a safety clone
You choose how each source is read: migrate a running VM while it stays in production, shut it down first for the cleanest copy, and optionally keep a full safety clone of the original.
- Online (recommended): the VM keeps running. For VMware, a snapshot freezes the disks and changed-block tracking ships only what changed, so downtime is limited to the cut-over
- Offline: the VM is shut down before the first copy and stays off - the cleanest copy, with downtime from the start
- Cut-over: optional graceful shutdown of the source, a final sync, checksum verification, then the new VM is powered on (or left off for you). The original is kept until you choose to delete it
- Optional safety clone (VMware): a thin, powered-off, network-disconnected copy with new MAC addresses, made before cut-over. Off by default; a running VM is copied crash-consistently with no downtime, and if the clone fails the migration does not start
- Destination space is checked before a copy starts, with a warning if it won't fit. Temporary snapshots and clones are removed after failed or cancelled jobs; a safety clone is only ever deleted by you
Scheduling & policies
Group VMs into policies with their own schedule, bandwidth limits, and retry behavior, instead of kicking off every migration by hand or writing your own cron jobs.
- Per-policy bandwidth caps so a migration doesn't starve production traffic
- Inventory refreshes run as a staggered background job, never on demand mid-wizard
Live performance dashboard
Throughput, jobs per hour, thin-copy, compression and deduplication savings, and job health across every migration and replica in flight, with history from real time out to 90 days - not just a progress bar.
- Average RPO and (once you've had a failover) average RTO, both from real measurements
- Per-job phase breakdown to the millisecond, so a slow run shows you exactly where the time went
- Live hypervisor latency in milliseconds, datastore and CPU counters, and any custom time range with drag-to-zoom
Disaster recovery mode
For VMware and OpenStack, keep replicating after cutover: continuous sync, failover, test failover and failback to Proxmox or another endpoint - licensed per protected seat, billed annually, running on the same install as your migrations.
- Test failover without disrupting the production replica or resetting its sync
- RPO tracked live against each policy's sync interval, with a breach alert
- Hyper-V to Hyper-V is protected a different way - Migration Center orchestrates Hyper-V's own native Replica directly (initial sync, test/planned/unplanned failover, reverse, health) - same licensing, no separate console
Role-based access & 2FA
Admin / operator / read-only roles, two-factor authentication via any TOTP app (Google Authenticator, Microsoft Authenticator, Duo, or similar), and a full audit log of who changed what.
- Repeated failed logins to a hypervisor stop automatically, so a changed password doesn't lock the account out
Encrypted at rest
Hypervisor credentials, SSH keys, and other secrets are encrypted at rest with rotatable keys - never stored in plain text, and only ever decrypted in memory for the duration of a job.
- License signing is handled by a separate internal service - never internet-facing
Down to the millisecond
See what a hypervisor is doing while a migration runs - not an hour later. Live latency every 5 seconds, the hypervisor's own datastore and CPU counters, and any time range from 10 seconds to 90 days.


Measured honestly
- Connect time (TCP + TLS) and one API round trip on an already-open session - no login per probe
- VMware datastore read / write latency and CPU; Hyper-V disk latency and CPU; Proxmox CPU and I/O wait (Proxmox has no latency counter, so none is invented); OpenStack reports none
- The chart keeps the worst value of each interval, so a spike is never averaged away
- Each job shows its own phases in milliseconds and the latency of its source and destination while it ran

Built for enterprise estates
Everything above runs the same on a small lab or a large data center. For bigger estates, Migration Center adds the scheduling, resilience and operational controls that long, busy migrations depend on.
Fair, predictable scheduling
Cut-overs and failovers always go first, then first copies, then routine syncs. Work is shared fairly between policies, limits protect every source host and destination node, and sync times are spread out so hundreds of VMs never all come due at once.
Scale out, not just up
Add worker capacity with a single command as the estate grows. A queue-wait view and an alert tell you before waits start to threaten your recovery targets - and the documentation includes a sizing guide.
Keeps going when things break
Restarts and upgrades let running jobs finish cleanly. A database outage does not lose results. An interrupted first copy continues where it stopped, and the part already copied is re-checked against the source.
Self-managing storage
Old job history and logs are cleaned up automatically, the appliance warns before its disk fills, and the destination is checked before a copy begins and watched while it runs.
Operations-ready
Health checks for every service, optional metrics for your monitoring system, support for a highly available PostgreSQL, and encrypted off-site copies of the appliance's own backups.
Secure by design
Role-based access, two-factor sign-in, a full audit trail, encrypted credentials and strict separation of tenants and data - the controls security teams ask for, in the box.
- Jobs are prioritised and limited per host, so production hypervisors are never swamped by migration traffic
- Throughput scales by adding workers; the documentation gives starting sizes for small, medium, large and enterprise deployments
- Designed so that no single failure - a restart, a lost connection, a database blip - silently loses work
See the recommended sizing, or talk to us about a large estate.
How it compares
Most tools in this space were built for one job and bolted the other on. Migration Center is built for the move to Proxmox specifically, with DR as a first-class second act - not the reverse.
| Migration Center | Zerto | Veeam | |
|---|---|---|---|
| Proxmox VE as a target | Core product - migration and DR | Not supported (VMware vSphere, Hyper-V and public clouds) | Yes - backup and restore since 2024; VM replication added in v13.1 |
| Moving VMs to Proxmox | Direct block-level copy with repeat syncs, then a short cut-over - no backup step | Not a Proxmox migration path | Yes - restore VMware / Hyper-V backups to Proxmox (a backup first; Instant Recovery added in v13.1) |
| Source platforms | VMware, Hyper-V, OpenStack (beta), Proxmox | VMware vSphere, Hyper-V, public clouds | The broadest list - VMware, Hyper-V, Proxmox, Nutanix AHV and more |
| Disaster recovery replication | Scheduled, change-tracked syncs (minutes) with failover and test failover | Continuous, journal-based, recovery points seconds apart - its core strength | Snapshot-based replication jobs |
| Cutting data sent over the wire | Changed-block tracking, empty-block skipping, optional zstd compression, and optional verified deduplication on the first copy to Proxmox - built in, no extra appliance | WAN compression and signature matching to reduce data sent over the WAN | Compression, plus WAN acceleration with global deduplication - needs a WAN accelerator pair and an Enterprise or Universal license |
| What you deploy | One appliance VM; nothing on hosts or in guests | A replication appliance on every host | A backup server plus a worker / proxy on hosts |
| Pricing model | One-time per VM, or annual per-seat DR | Per-VM subscription, quote-based | Per-workload subscription, tiered editions |
Zerto and Veeam are trademarks of their respective owners; this comparison reflects their publicly documented capabilities as of November 2026 and isn't endorsed by either company. Veeam's support for Proxmox is documented here and its supported workloads here; HPE lists Zerto's supported platforms here. Vendors change their products - check their current documentation, and if something here is out of date, let us know and we'll correct it.
See it on your own workload
Download it free - it runs in evaluation mode with no license required (30 days, up to 5 VMs / 750 GB for migration, 3 VMs for replication - see the limits), so you can try a real migration before buying. The download is a ready-to-run appliance image of about 7 GB.