Comparison

Managed vs self-hosted Proxmox Backup Server

A factual comparison from a team that has operated both, for over a decade. No marketing fluff. Costs, trade-offs, and the cases where each approach wins.

A self-hosted Proxmox Backup Server is the right choice when your team already has on-call coverage for backup infrastructure, spare rack space, and the time to keep Debian and PBS patched. For everyone else, a managed service reaches first successful backup faster, lowers total cost over three years, and removes the single-site risk of keeping your backup server in the same rack as the production it protects.

Start 7-day free trial See managed pricing

Quick verdict

If you are reading this page, you probably fall into one of three profiles. Here is the honest answer for each:

Homelab enthusiast with time to tinker

Self-host, if you enjoy it

PBS is a well-documented upstream project. Running it on a second-hand machine with ZFS and 8 GB of RAM is educational and cheap. The only real risk is that your only backup lives in the same apartment as the VMs it protects.

SMB without a dedicated sysadmin team

Managed, without hesitation

The operational overhead of a PBS host (patching, monitoring, disk replacement, verify failures) is a direct distraction from your business. A managed plan at €6 to €60 per month covers everything you would have otherwise done in your evenings.

Enterprise with a 24/7 ops team

Usually managed for off-site, possibly self-hosted for on-site

The 3-2-1 rule requires at least one off-site copy. Self-hosting that off-site copy means running a second datacenter you pay for anyway. Managed is cheaper and removes the correlated-failure risk of a single vendor or site.

Detailed comparison

Eight dimensions that come up in every evaluation call we do. We have tried to state each point in terms that can be measured, not marketed.

Dimension Self-hosted Cloud-PBS (managed)
Time to first successful backup 4-10 days (hardware procurement, racking, Debian and PBS install, TLS setup, first verify) Under 10 minutes (register, paste the repository string into Proxmox VE, run the first job)
Hardware CAPEX €1,500 to €4,000 for a decent PBS host (server, RAM, disks, network). Renew every 4-5 years. Zero. Operating expense only.
Monthly operational cost Hardware amortised (~€50-100/mo) plus power, bandwidth, rack space, and 2-4 hours of sysadmin time per month. €6 to €180 per month depending on storage. Zero sysadmin time.
Patching responsibility You patch Debian, the kernel, PBS, ZFS tooling. Every CVE is yours. Included. We patch during a maintenance window we announce in advance.
Monitoring You set up Prometheus or Zabbix, write alert rules, handle pager rotations. 24/7 monitoring with automatic failover and proactive customer notification.
Disk failures You order a disk, schedule the replacement, wait for the pool to rebuild. Transparent. You never know a disk failed.
Off-site guarantee Only if you run a second host in another site. A single PBS in the same rack as the PVE is not a backup. Off-site by construction. Different datacenter, different AS, different jurisdiction if you choose.
Expertise if something breaks Your team, Google, the Proxmox forum. Escalation paths are what you can find on a Wednesday night. Direct access to engineers who operate hundreds of PBS instances. Acknowledgement in one hour, substantive response in four.

Three-year cost breakdown

A worked example for a mid-sized fleet of 10 VMs totalling 5 TB of backup data, with daily backups kept 30 days and weekly kept 12 months. Numbers are rounded and ignore inflation.

Line itemSelf-hostedCloud-PBS (Shared 5 TB)
Server hardware (1 unit, 5-year amortisation, used over 3 yr)€2,100€0
Rack space and power (€40/mo)€1,440€0
Bandwidth to off-site target€360Included
Sysadmin time (3 h/mo at €60/h internal loaded)€6,480€0
PBS software licence€0 (open source)€0
Cloud-PBS subscription (€60/mo for 5 TB)€0€2,160
Incident recovery (1 disk replacement + 1 restore drill)€600€0
Total over 3 years€10,980€2,160

Self-hosting can be cheaper if you remove the sysadmin line entirely, which is only realistic when a person is already paid to cover that scope. For anyone who has to buy that person's time, managed wins on cost alone.

Run the numbers on your own workload

When self-hosting still wins

Three real cases where we recommend against managed hosting for the PBS layer:

  • You are already running PBS and your team has the skill and appetite. Moving off it because "managed is fashionable" is a poor reason.
  • You operate a sovereign environment where backup data may not leave a specific air-gap or compliance boundary. Some defence, critical infrastructure and healthcare contexts fall here.
  • You need custom PBS tuning outside the supported configuration surface, for research or high-experimental workloads. Managed services trade flexibility for operability.

When managed is the better choice

The five cases where managed wins clearly:

  • Your primary concern is business continuity, not infrastructure craft.
  • Backup is not a differentiator for your product or service. Time spent on it is overhead.
  • You want an immediate off-site copy without building a second site.
  • You care about total cost of ownership over three years, not just the headline hardware price.
  • You lack 24/7 coverage for backup infrastructure failures. Managed gives you coverage you would otherwise have to hire.

Moving from self-hosted to managed

You do not have to throw away your existing setup. Two migration paths work today:

Push-then-sync

Keep your self-hosted PBS. Configure Cloud-PBS as a remote on it and schedule a pull or push sync. For a week or two, both datastores run in parallel. When you are confident the managed copy is complete and verified, point new PVE backup jobs at the managed datastore and decommission the local one.

Physical seed

For datasets larger than what your uplink can move in a reasonable window, we accept physical seed drives by registered post. We ingest the data on arrival, preserving deduplication state, and the datastore is ready as if it had been native from day one.

Managed vs self-hosted PBS: frequently asked questions

Does managed PBS actually cost less than self-hosted?
Yes, once you price a sysadmin's time realistically. Hardware and power alone make self-hosting look cheaper, but the two to four hours a month of operational work during normal conditions — plus the occasional disk replacement, PBS upgrade or verify-failure investigation — crosses into the three-year TCO. A fleet that backs up 5 TB pays about €2,160 over three years on a managed plan. Self-hosting the same workload with realistic loaded internal costs runs over €10,000.
Is self-hosted PBS more secure than managed?
No, and often the opposite. Both run the same upstream PBS with the same AES-256-GCM client-side encryption. Managed providers apply patches faster than most internal teams because it is their only job. The perceived security advantage of self-hosting is the air-gap — but the moment you need an off-site copy, which every backup strategy requires, the air-gap is gone.
Can I get vendor lock-in with a managed PBS?
Not if it runs upstream PBS, which Cloud-PBS does. The datastore format is the standard PBS format. You can copy your backups to any other PBS instance, self-hosted or managed elsewhere, using the same remote sync tools. The exit cost is close to zero.
What happens to my backups if the managed provider shuts down?
You should ask this of any vendor. With Cloud-PBS specifically, the commercial terms include an orderly exit window during which you can push or pull your full dataset to any PBS of your choice. Because backups are stored in upstream PBS format, you do not depend on proprietary tooling to recover them.
Is it reasonable to run self-hosted PBS on a Synology or QNAP?
PBS on a NAS is a common pattern. It works for small fleets but you hit limits around deduplication performance (PBS wants fast CPU and RAM), ZFS stability (most NAS filesystems are not ZFS), and PBS upgrade cadence (you are at the mercy of the NAS vendor's Linux kernel). For production workloads past 1 TB we usually recommend either a proper PBS box or a managed instance.
How do I justify managed to a finance team?
Build a 3-year TCO table with three columns: hardware and power, sysadmin hours at loaded cost, managed subscription. The managed column is almost always the smallest. If your finance team insists on CAPEX over OPEX, that is a business policy conversation, not a technical one.

Try Cloud-PBS on your workload

Create a managed datastore, point a single Proxmox VE backup job at it, and compare throughput and cost on your own numbers. The trial is free for 7 days.

Start free trial Read the hub guide