Latest
Storage

NAS, USB Drive, or Proxmox Backup Server?

Compare a NAS share, external USB drive, and Proxmox Backup Server by failure separation, maintenance, scale, and recovery workflow.

By Lab Gremlin · July 30, 2026

A Proxmox backup destination has two jobs. It must accept backups reliably, and it must still be usable when the virtualization host has failed. A NAS, an external USB drive, and Proxmox Backup Server can all meet the first requirement. They differ sharply on the second.

The practical choice depends on how many Proxmox nodes you protect, how often backups run, how much administration you will accept, and whether the destination shares power, storage, or credentials with the systems it protects. This comparison uses documented capabilities rather than product testing.

The three choices in one table

Destination Best fit Main advantage Main limitation
External USB drive One small node and a limited backup set Low setup cost and simple file-based restores Easy to leave attached, forget, fill, or lose with the host
NAS share Several systems already using maintained network storage Central capacity with NFS or CIFS access Backup features depend on the NAS and the Proxmox VE file-backup workflow
Proxmox Backup Server Recurring backups, multiple guests, or more than one node Incremental transfer, deduplication, verification, pruning, and sync workflows Requires another maintained system and a suitable datastore

None of these destinations becomes a complete backup plan by itself. Capacity, retention, notifications, a separate failure domain, and a restore test still have to be designed.

External USB drive: the smallest workable starting point

A directly attached drive can be reasonable for a single Proxmox node when the budget and data set are small. Proxmox VE can write backup archives to a directory storage configured for backup content. The archive contains the guest data and configuration needed by the integrated restore workflow. This is a straightforward model: the host creates files, and the operator copies or rotates them.

The weak point is operational. A drive that remains plugged into the same host shares several hazards with it. A power event, accidental deletion, compromised credentials, theft, or a failed restore path can affect both the source and its backup. Disconnecting and rotating drives improves separation, but only when the schedule is followed and missed jobs are visible.

Check the filesystem, mount behavior, capacity alerting, and device identity before scheduling jobs. A mount point that silently falls back to the host filesystem can send backups to the wrong disk. A portable drive powered only by USB also adds cable and power-delivery variables. Use a stable mount by device identifier, restrict the storage to backup content, and confirm the target path before the first scheduled run.

This route makes the most sense when someone will rotate the media, store one copy away from the host, and periodically restore a guest from it. If those habits will not happen, the low purchase cost does not translate into a dependable recovery plan.

NAS share: simple central storage over the network

A NAS can expose an NFS or CIFS share that Proxmox VE treats as file-level storage. The official Proxmox VE administration guide lists both backends as able to hold backup content. That makes a NAS useful when it already has monitored disks, sufficient capacity, access controls, and a maintenance routine.

Network storage separates the backup data from a failed Proxmox boot disk and can serve several nodes. It does not automatically separate every failure domain. A NAS and virtualization host on the same UPS, switch, room, or administrative account can still fail together. RAID can keep a NAS running after a disk failure, but RAID is not another backup copy.

Plan for the network path as carefully as the disks. Confirm that the share reconnects after a NAS or switch restart, that only the intended nodes can write to it, and that Proxmox reports a failed job when the share is unavailable. Avoid putting active guest disks and their only backups in the same storage pool. Capacity alerts should account for retention growth rather than only today’s backup size.

A NAS is often the middle option: easier to centralize than rotated USB media and simpler to introduce than a dedicated backup service. It remains a file-backup destination, so integrity checks, off-site copies, and restore exercises need separate procedures unless the NAS provides and maintains them.

Proxmox Backup Server: more backup control, more system to maintain

Proxmox Backup Server is a dedicated backup platform rather than only a share. Proxmox documents incremental transfers to the server with server-side deduplication. Each snapshot still references the data required for a full backup. Its datastore workflows also include pruning, garbage collection, verification jobs, namespaces, access roles, and synchronization to another server.

Verification is a meaningful distinction. Proxmox Backup Server records checksums in a backup manifest and can run jobs that verify stored data. That can find damaged data before an emergency, although verification is not a substitute for restoring and booting a guest. The current documentation specifically recommends regular restore testing to a new guest.

The tradeoff is another machine, operating system, update path, datastore, and set of credentials. Proxmox’s production guidance calls for at least 4 GiB of memory for the operating system, filesystem cache, and backup daemons, plus additional memory as storage grows. It also favors storage suited to random I/O. Those requirements make a very small leftover computer possible for evaluation, but not automatically appropriate for a busy production datastore.

A dedicated physical system creates clearer separation from a Proxmox host than installing the backup server as a guest on that same host. A virtualized backup server can still be useful in a lab, but recovery becomes more complicated if the failed platform must be restored before the backup service can start.

Compare failure domains before feature lists

Write down what each option shares with the protected node:

  • Power: Does one outlet, UPS, or power supply affect both copies?
  • Storage hardware: Are active guests and backups on the same disks or enclosure?
  • Network: Can a failed switch or bad configuration block recovery?
  • Credentials: Can an account used by the host delete existing backups?
  • Location: Would theft, water, fire, or physical damage reach both systems?
  • Administration: Does one mistaken retention or deletion action affect every copy?

Proxmox Backup Server supports narrow access roles and remote synchronization, which can reduce some of these risks. A NAS may offer snapshots, immutable storage, or replication, but those capabilities are model-specific and should be verified in its documentation. A disconnected USB rotation can provide physical isolation, but only after the backup completes and the drive is actually moved.

Choose by workload and operating habit

Choose a USB drive when one modest node needs a low-cost starting point and manual media rotation is realistic. Keep at least one recent copy disconnected and away from the host.

Choose a NAS share when a maintained NAS already exists, several systems need central capacity, and file-based Proxmox backups fit the recovery plan. Give backup data its own permissions and retention budget.

Choose Proxmox Backup Server when backups run frequently, several guests or nodes need retention, and verification plus deduplication will reduce recurring work. It is also the stronger base when a second backup server will receive synchronized copies.

A common growth path is USB for the first node, a NAS when storage becomes shared, and Proxmox Backup Server when backup administration becomes repetitive. Migration is not mandatory. The right point to change is when the current process is being skipped, cannot meet the recovery window, or gives too little warning when something breaks.

A narrow shopping path for the USB option

Before buying storage, estimate retained backup data, expected change between backups, retention count, and growth. Then add room for maintenance rather than filling the destination immediately. Confirm the connector, cable length, power source, supported filesystem, warranty terms, and whether the enclosure reports drive health. Do not select a destination only from the capacity printed on its box.

This article contains affiliate links. As an Amazon Associate, LabGremlin earns from qualifying purchases.

If the USB route fits the plan, this Amazon search for desktop external USB hard drives is a starting point for comparing form factor and capacity. It is not a product endorsement. Verify the exact drive, enclosure, filesystem plan, power requirements, and return terms before purchasing. A search result does not establish durability or compatibility.

For a NAS or dedicated backup server, begin with workload and failure-domain requirements before shopping for an enclosure. The existing Proxmox storage planning guide explains why active storage, snapshots, and backups should be treated as separate roles. If a second physical machine is under consideration, the mini PC versus used business desktop comparison can help frame expansion and serviceability.

Implement the destination, then prove recovery

  1. Configure the destination for backup content and restrict its permissions.
  2. Run one manual backup and confirm its size, duration, and completion status.
  3. Set retention that fits measured capacity and recovery needs.
  4. Configure failure notifications and confirm that they arrive.
  5. Restore a noncritical guest to a new ID and isolated network.
  6. Record the destination, credentials recovery method, retention, and last successful restore.
  7. Add another copy with different credentials or a different location when the data matters.

The destination decision is complete only when a restore works without depending on the failed host. A modest device with a rehearsed process is more useful than a sophisticated target that has never been read back.

Sources and research date

Primary sources and the product-search destination were reviewed July 30, 2026. This comparison describes documented capabilities and planning tradeoffs. It does not claim LabGremlin testing or ownership of any destination.