Proxmox VE and Vates can both host a Windows/Linux VM estate coming from VMware. The choice turns on retained backup software, storage, daily operations and support. Here are the differences that materially change the decision.
The five decisive differences
- Existing backup: Veeam and NAKIVO provide different integrations across the targets; keeping the software does not guarantee keeping every feature.
- Storage: Proxmox integrates ZFS/Ceph; Vates offers optional XOSTOR. This matters less with a retained SAN.
- Operations: VE manages its cluster natively; Xen Orchestra groups pools, backup, replication and maintenance.
- Specific workloads: LXC and shared NVIDIA vGPU can favour Proxmox; passing an entire GPU through addresses a different requirement.
- Price: VE is charged per socket, Vates notably per host. PBS remains optional: the budget changes with the chosen backup solution.
Our recommendation for your situation
We favour Proxmox for its storage integrations and specific workloads, Vates for workflows grouped around XO. These illustrative cases refine that choice; they are not customer results.
| Situation | Our preferred choice | What could change that choice |
|---|---|---|
| Retained SAN and Veeam, integrated application protection required | Proxmox VE, based on the plug-in features compared in this article | Veeam version, application matrix and required restores; XCP-ng remains a candidate if its coverage is sufficient or separate application protection is planned |
| New hyperconverged infrastructure | Proxmox/Ceph when Ceph shapes the target; Vates/XOSTOR when block replication and XO workflows fit better | Pool size, large disks, failure reserve, network and operating skills; cost the XOSTOR option |
| Small three-host estate | VE Basic/Standard for a controlled platform requirement; Vates Essential/Essential+ for the appropriate XO bundle | Single/dual-socket hosts, XO features and support commitments; the price table can change the ranking |
| Grouped management, backup and replication | Vates/XO, if grouping actually reduces the tools operated | Already-standardised third-party backup, remote-site options, functions still in XO 5 |
| Shared NVIDIA vGPU, LXC or ZFS required | Proxmox VE, for the corresponding integrations | NVIDIA entitlements, validated hardware and container protection; the Veeam plug-in does not cover LXC |
| Critical infrastructure across sites | No default winner: compare recovery architectures and Premium/Enterprise offers | RPO/RTO, site failure, management and key access, network and spare capacity; 24/7 support does not guarantee application recovery |
The following sections detail the functions, versions and conditions behind these recommendations.
Prices: five hosts with your existing backup retained
If your software covers the target and required restores, no PBS subscription is added. For five hosts, retaining storage and without a secondary site:
| Annual scenario, excluding retained backup software | Proxmox VE, without PBS | Vates VMS with XOA |
|---|---|---|
| Standard / Pro — 5 single-socket hosts | 5 × €550; platform total: €2,750/year | 5 × €1,000; platform total: €5,000/year |
| Standard / Pro — 5 dual-socket hosts | 10 × €550; platform total: €5,500/year | 5 × €1,000; platform total: €5,000/year |
| Premium / Enterprise — 5 single-socket hosts | 5 × €1,100; platform total: €5,500/year | 5 × €1,800; platform total: €9,000/year |
| Premium / Enterprise — 5 dual-socket hosts | 10 × €1,100; platform total: €11,000/year | 5 × €1,800; platform total: €9,000/year |
Public annual prices excluding tax, verified on 30 October 2026, without discounts. Support commitments differ between these offers. Retained backup software costs must be confirmed on both sides; they are assumed neither zero nor unchanged. XOA remains included with Vates even when its backup features are unused.
Within platform subscriptions alone, Proxmox costs less with single-socket hosts, Vates with dual-socket hosts. With new vendor-supported native backup, the first scenario becomes €4,990/year for VE + PBS versus €5,000/year for Vates. Other budgets and support levels appear below.
Sources: VE, PBS, Vates VMS and XOSTOR.
What we are comparing
Proxmox VE combines KVM for VMs and LXC for system containers. PBS adds backup; PDM centralises multiple environments. Vates VMS combines Xen-based XCP-ng and Xen Orchestra for management and backup. XOSTOR and XO Proxy are distributed-storage and remote-backup options.
Sources: Proxmox VE, PDM, Vates VMS.
DRS, vMotion, HA and vSAN: VMware counterparts
This table distinguishes an available feature from equivalent behaviour. The Proxmox dynamic balancing described here applies to VE 9.2 and HA-managed resources.
| VMware feature | Proxmox VE / PBS / PDM | Vates: XCP-ng / Xen Orchestra | Key distinction |
|---|---|---|---|
| vCenter | Cluster management in VE; PDM across environments | XO centralises hosts/pools and backups | Both offer central management, with a different distribution of roles |
| DRS: placement and balancing | VE 9.2 CRS and Dynamic Load Balancer for HA-managed resources | XO load-balancer plugin with plans and VM migrations | Both balance load; Proxmox's scope is HA resources, while an XO plan selects pools and eligible VMs |
| Affinity / anti-affinity | HA rules between resources and nodes, including strict constraints | Tag-based affinity/anti-affinity in XO plans | Strict HA constraints and XO's best-possible distribution have different semantics |
| vMotion | QEMU/KVM live migration | XCP-ng live migration through XO | Available on both sides, with platform-specific CPU, device and network constraints |
| Storage vMotion | Online disk moves within supported storage combinations | Live storage migration to another SR, including VM migration where applicable | Both provide storage mobility; distinguish disk moves from moving a VM together with its disks |
| DPM: consolidation and power | Documented CRS balances HA resources; automatic host shutdown/wake-up is not established by that mechanism | Density evacuates and shuts down compatible hosts; Performance can power them back on | XO has the documented energy workflow, subject to power-on configuration and VM mobility |
| vSphere HA: host failure | HA resource recovery, quorum and watchdog fencing | Pool HA, network/storage heartbeat and protected-VM restart | Both recover through restart, using different failure-detection designs |
| Admission Control / spare capacity | Spare capacity must be sized; CRS placement rules alone do not establish identical vSphere admission policy | Pool capacity calculation and tolerated host-failure setting | N+1 capacity and admission policy are distinct from a balancing threshold |
| VM Monitoring | HA resource monitoring; a hung guest OS needs separate treatment | Host/pool HA; guest OS health needs separate treatment | Host-failure recovery does not establish VMware Tools heartbeat equivalence |
| Fault Tolerance | The described HA restarts the VM | The described HA restarts the VM | Neither described HA mechanism is a synchronised secondary VM like VMware FT; that requirement needs another continuity solution |
| vSAN | Integrated Ceph; SAN/NFS remain options | Optional XOSTOR; SAN/NFS remain options | Different distributed engines; do not assume automatic translation of vSAN objects and policies |
| NSX | VE SDN, VXLAN/EVPN and firewall capabilities | Open vSwitch, VLANs and overlays through XO's SDN controller | Both provide useful networking; no complete equivalence to all NSX capabilities is claimed |
| SRM / disaster recovery | PBS, remote copies and recovery procedures to organise | XO backup and replication to another storage/pool | Protection and replication exist; complete DR orchestration still needs to be designed |
Sources: VE 9.2 dynamic balancing, Proxmox HA and rules documentation, QEMU migration/disks, XO plans, affinity and power, XCP-ng migration, XCP-ng HA and capacity, VMware FT definition.
Retaining Veeam or NAKIVO: the useful differences
The backup server and plug-ins must cover the hypervisor version. These integrations are documented as of 28 October 2026.
| Software or requirement | Proxmox VE | Vates / XCP-ng |
|---|---|---|
| Veeam, hypervisor integration | Proxmox plug-in; the current matrix covers VE 8.2–9.2 on x86-64 installed from the official ISO | Xen plug-in introduced with VBR 13.1; the current requirements page specifies 13.1.1.18 or later and covers XCP-ng 8.3 or later |
| Veeam, consistency and restore | Application-aware processing and application-item recovery documented with QEMU Guest Agent installed and enabled before backup; LXC excluded from the plug-in | VM and file restore documented; the current plug-in supports neither guest quiescence nor VM replication; do not infer identical application protection |
| NAKIVO, published matrix | VE 8.x–9.2 listed; for VE 9, documentation notably requires a VM machine type of 10 or lower | XCP-ng is not listed as a hypervisor-backup platform on the consulted page; agent-based protection must be confirmed for the relevant OS and contract |
| In-guest agent | Can complement application protection or cover uses outside the plug-in, depending on the software | Same possibility depending on software; protecting an OS with an agent does not establish full VM and configuration recovery |
For application protection integrated into a Veeam job, the Proxmox integration is more complete here. XCP-ng remains a candidate for VM/file recovery with the required Veeam version. The consulted NAKIVO matrix favours Proxmox for retaining explicitly listed hypervisor integration.
Cost the version, usage rights and workers. Jobs, history and per-socket VMware licences are not assumed to transfer automatically: verify an old restore point and a new job on the target.
Sources: Veeam / Proxmox requirements, Proxmox limitations, Xen plug-in introduction, current Xen requirements, Xen restore, Xen limitations, NAKIVO matrix.
Administration: native VE clusters or grouped XO management
VE is administered through its nodes, without an additional cluster-management appliance. PDM adds inventory, monitoring and operations across clusters or sites, with local interfaces for detailed settings.
XO centralises XCP-ng pools and their backups. XOA groups roles distributed across VE, PBS and potentially PDM on the Proxmox side. VE suits native cluster management; XO gains an advantage when management and protection should share one tool.
Sources: VE management, PDM architecture, Vates VMS, plans and components.
New backup: the PBS model versus XO
PBS uses a deduplicated chunk repository, referenced by each restore point's indexes. Transfers are incremental, with dedicated verification and maintenance operations.
XO incremental backup combines an initial backup with deltas, reconstructed for restore and merged according to retention. XO distinguishes backup repositories from replication to VM storage prepared for recovery.
We favour PBS for its dedicated deduplicated repository, XO for grouped backup/replication/management. Operations require maintaining the PBS service on one side, XO chains, reference snapshots and merges on the other. These models alone do not rank restore speed or reliability.
Sources: PBS technical architecture, XO incremental backup, protection types.
Storage: ZFS/Ceph or XOSTOR
Proxmox integrates ZFS for local storage and Ceph for distributed storage, including RBD for block and CephFS for files. XOSTOR combines DRBD and LINSTOR within an XCP-ng pool: at least three hosts, appropriate networking and a documented maximum of seven hosts per XOSTOR pool.
Proxmox is the direct choice for ZFS or a Ceph-based target, particularly beyond that XOSTOR limit. Vates remains a coherent block-replication choice within its prerequisites. Both also accept SAN/NFS: retaining an array reduces the Ceph/XOSTOR distinction's weight.
Sources: VE storage and Ceph, XOSTOR architecture and prerequisites.
Hardware, large disks and GPUs
| Requirement | Proxmox VE | Vates / XCP-ng |
|---|---|---|
| Disk exceeding 2 TiB | Raw/QCOW2 formats and block storage depending on backend; no dependency on VHD | Production QCOW2 support on updated 8.3, up to 16,381 GiB; VHD remains limited to 2,040 GiB. The FAQ excludes QCOW2 on LINSTOR and SMB SRs |
| Shared NVIDIA vGPU | VE is a supported NVIDIA vGPU hypervisor; support requires VE Basic/Standard/Premium and the corresponding NVIDIA entitlement | XCP-ng documentation states NVIDIA vGPU is unsupported; passing an entire card through remains possible |
| PCI/GPU dedicated to a VM | Passthrough available with IOMMU; a host-bound resource can prevent live evacuation | Passthrough available; migration capabilities depend on GPU hardware and software, not simply the presence of passthrough |
| Retained servers and array | Driver and controller compatibility must be established; Proxmox advises against hardware RAID underneath ZFS/Ceph | Vates HCL and alternate drivers documented; hardware outside the HCL can be supported with a different validation level |
A VM with a 4 TiB disk no longer rules out XCP-ng: storage and backup must support the selected QCOW2 combination. Shared NVIDIA vGPU favours Proxmox; passing an entire card through does not provide that sharing.
vCPU/RAM limits also depend on OS, firmware and memory snapshots. The XCP-ng matrix notably distinguishes Windows; its maximums are not production sizing. ERP, database and appliance support must also be confirmed with their vendors.
Sources: QEMU formats and migration, XCP-ng QCOW2 FAQ, NVIDIA vGPU on VE, XCP-ng GPU, GPU migration constraints, VE hardware requirements, XCP-ng hardware support, XCP-ng matrix.
Networking: SDN on both sides, integrated EVPN in Proxmox
Proxmox documents VLANs, VXLAN and EVPN/BGP. XCP-ng uses Open vSwitch, with VLANs, bonding and global private networks through XO's SDN controller using GRE/VXLAN.
A platform-managed EVPN target favours Proxmox's integration. With retained VLANs and network services, Vates provides the usual primitives. NSX microsegmentation, routing and automation still require specific design.
Sources: Proxmox SDN, XCP-ng networking.
DRS and HA: scope makes the difference
VE 9.2 CRS uses real utilisation for placement; the Dynamic Load Balancer automatically migrates HA guests according to their rules.
In XO, Performance spreads load, Density consolidates and can shut down eligible evacuated hosts, and Simple applies affinities without active balancing. Performance can start hosts with a configured power-on mode.
The decisive distinction concerns constraints: Proxmox HA affinity can be strict; XO uses best-possible tag-based distribution when there are not enough hosts to separate every VM. A mandatory rule during failure therefore deserves particular attention.
For host recovery, Proxmox uses a watchdog and XCP-ng network/storage heartbeats. The VMware table distinguishes this HA from hung-OS monitoring and VMware FT.
Sources: VE 9.2 announcement, XO plans and conditions, Proxmox HA, XCP-ng HA.
Maintenance and lifecycle
| Operation | Proxmox VE | Vates / XCP-ng and XO |
|---|---|---|
| Patches and reboots | APT updates through UI or CLI; HA maintenance mode moves managed resources before node-level intervention | Rolling Pool Update: VM moves, patches and reboots orchestrated through XO, subject to mobility and shared-storage requirements |
| Major version change | Debian base and VE components upgraded; Ceph also has its own upgrade procedure | XCP-ng recommends the ISO procedure; XAPI state is retained, but not all system customisations survive. A patch RPU does not replace this procedure |
| Management unavailable | Cluster administration accessible through other nodes; losing PDM is distinct from losing hypervisors | XO is a separate service: its unavailability interrupts its workflows without itself constituting VM failure. XO configuration and pool metadata require backups |
| Interfaces and support horizon | VE interface for clusters, PDM for multiple environments; track versions and Ceph compatibility together | XCP-ng 8.3 LTS supported until 30 November 2028; XO 6 and XO 5 coexist, with some operations remaining in XO 5 |
XO orchestrates pool maintenance; Proxmox retains control at each node. The consulted RPU procedure temporarily disables HA, XO balancing and backup jobs, then re-enables them. Availability during that window must be planned.
Immovable VMs or insufficient spare capacity can require interruption. Ceph/XOSTOR also need sufficient capacity and copies throughout maintenance and rebuild.
Sources: VE updates, HA maintenance, Ceph management, RPU and conditions, XCP-ng upgrades, XCP-ng lifecycle, XO 5 / XO 6 split, metadata backup, XO architecture.
Security, delegation and automation
| Requirement | Proxmox | Vates / Xen Orchestra |
|---|---|---|
| Authentication | LDAP/AD, OpenID Connect and second factors including TOTP/WebAuthn | LDAP/OIDC providers and TOTP documented; the login path depends on provider and interface |
| Delegation | Object-level roles and permissions; API tokens with their own rights and expiry | Historical XO 5 ACLs and RBAC for REST/XO 6; verify UI coverage rather than inferring it from API coverage |
| Traceability | Access and task logs to centralise according to retention requirements | Documented Audit Log plug-in, disabled by default; activation and retention must be organised |
| Automation | VE REST API and tokens; PBS/PDM calls target their own services | XO REST API with tokens and embedded OpenAPI reference; pool administration through the same service |
Proxmox protected status is not storage-enforced WORM. XO local immutability and S3 Object Lock are distinct; root on the repository can lift local protection.
Across both architectures, separating backup accounts, storage administration and encryption keys matters more than a global KVM/Xen ranking. PBS supports copying to another server; permissions and deletion propagation need explicit configuration.
Sources: VE identities, permissions and tokens, VE logging, XO users and audit, XO RBAC, XO REST API, Proxmox backup protection, XO immutability, synchronisation to another server.
Migration: VE import or XO pre-copy
Proxmox includes an ESXi import wizard, mapping part of the VM configuration. XO documents pre-copy while the VMware VM remains active, followed by final cutover; the guide still uses XO 5 for that import.
Integrated import simplifies the VE workflow; XO pre-copy prepares large volumes before shutdown. Source storage, disks, drivers and bandwidth affect duration. On both paths, returning to VMware must address writes made after cutover.
Sources: Proxmox wizard, vendor announcement, XO VMware migration.
Budgets and support: the details
The first table's pricing assumptions apply here: annual euros excluding tax, public prices as of 30 October 2026, no discounts, retained storage and no secondary site.
Five hosts with new vendor-supported native backup
One subscribed PBS server is selected on the Proxmox side; XO is included with Vates. These totals apply only to that backup choice.
| Annual scenario | Proxmox VE + PBS | Vates VMS with XOA |
|---|---|---|
| Standard / Pro — 5 single-socket hosts | VE: 5 × 550 = €2,750; PBS: €2,240; total: €4,990/year | 5 × €1,000; total: €5,000/year |
| Standard / Pro — 5 dual-socket hosts | VE: 10 × 550 = €5,500; PBS: €2,240; total: €7,740/year | 5 × €1,000; total: €5,000/year |
| Premium / Enterprise — 5 single-socket hosts | VE: 5 × 1,100 = €5,500; PBS: €4,480; total: €9,980/year | 5 × €1,800; total: €9,000/year |
| Premium / Enterprise — 5 dual-socket hosts | VE: 10 × 1,100 = €11,000; PBS: €4,480; total: €15,480/year | 5 × €1,800; total: €9,000/year |
The first scenario is €10 apart; Vates is lower in the other three. Five hosts meet Vates Pro's three-host and Enterprise's four-host minimums.
Three hosts: Basic, Standard, Essential or Essential+
VE Basic: €370/socket/year, three tickets, one-business-day first response, unchanged VE features. Vates Essential: €2,000/year for up to three hosts, six tickets; Essential+: €4,000/year, unlimited tickets and full features. Both include XOA and business-day support.
| Three hosts, annual platform budget | Proxmox VE Basic / Standard | Vates Essential / Essential+ |
|---|---|---|
| Single-socket | Basic: 3 × €370 = €1,110; Standard: 3 × €550 = €1,650 | Essential: €2,000; Essential+: €4,000 |
| Dual-socket | Basic: 6 × €370 = €2,220; Standard: 6 × €550 = €3,300 | Essential: €2,000; Essential+: €4,000 |
These budgets do not add PBS. Three hosts also permit Vates Pro; required XO features and support determine the plan.
Charging units and options
VE is subscribed per socket; PBS separately per server. PBS can also run without a subscription. XOA means an appliance, not backup disk capacity.
| Item | Proxmox | Vates |
|---|---|---|
| Virtualisation, Standard / Pro tier | VE Standard: €550/socket/year | VMS Pro: €1,000/host/year, one XOA included |
| Virtualisation, Premium / Enterprise tier | VE Premium: €1,100/socket/year | VMS Enterprise: €1,800/host/year, one XOA included |
| Supported native backup if selected, Standard / Pro tier | Optional PBS Standard: +€2,240/PBS server/year | XOA included in Pro, with that plan's features |
| Supported native backup if selected, Premium / Enterprise tier | Optional PBS Premium: +€4,480/PBS server/year | XOA included in Enterprise, with that plan's features |
| Multi-environment management | PDM included with eligible subscriptions | One XO appliance included in the cited offers |
| Distributed storage, Standard / Pro tier | Ceph integrated into VE; no separate Ceph subscription line in this calculation | Optional XOSTOR Pro: +€800/host/year |
| Distributed storage, Premium / Enterprise tier | Same Ceph principle | Optional XOSTOR Enterprise: +€1,200/host/year |
For five hosts, XOSTOR adds €4,000/year with Pro or €6,000/year with Enterprise: Vates totals become €9,000 or €15,000/year. No additional Ceph subscription line is added to VE in this calculation. Hardware, disks and networking still need funding.
Different support commitments
As of 30 October, Proxmox Standard awaits the 24/7 coverage announced for the end of 2026; Vates Pro covers business days. Proxmox Premium and Vates Enterprise advertise 24/7. Critical first response: 4 hours / 2 hours for Proxmox Standard / Premium, versus 24 hours / 1 hour for Vates Pro / Enterprise, under their respective conditions and severities. These are not guaranteed resolution times; quotas and included services differ.
Three years and costs beyond subscriptions
For five single-socket hosts at Standard/Pro level, at constant annual prices without discounts, platforms total €8,250 for VE versus €15,000 for Vates over three years. With PBS Standard: €14,970 versus €15,000. Vates commitment discounts are not applied; these projections do not guarantee future prices.
Total cost adds migration, training, operations, hardware, backup capacity, overlapping coverage during transition and adaptations to retained licences. Another PBS, XOA, XO Proxy or additional options also change the budget. These tables are not a quotation.
Sources: VE, PBS, Vates VMS and XOSTOR.
Validate the selected target
The choice starts with retained components, functions used and operating cost. A focused POC then confirms a representative migration, a restore, a host failure and maintenance.
VSHIFT supports both solutions: migration to Proxmox, migration to Vates or a discussion about your VMware exit platform.
Sources and scope
Official sources consulted on 27 and 28 October 2026, linked within each section. Current documentation, notably Proxmox's documentation branch, must be checked against deployed versions; XO interfaces, features and commercial rights also depend on the selected plan. Recommendations are our architectural assessment, without a measured performance, reliability or security ranking.
Sources: Proxmox VE, VE features, PDM, PBS architecture, Vates VMS, XO backup, XCP-ng protection, XOSTOR, XCP-ng networking, XCP-ng HA, XO balancing, Proxmox import, initial announcement, XO migration, Proxmox prices, Vates prices.