Ask any administrator who switched to Proxmox Backup Server what impressed them first, and the answer is almost always the same: the storage numbers. Backups that used to eat terabytes suddenly fit into a few hundred gigabytes, and the growth curve flattens almost immediately after the first run. This happens thanks to deduplication — the core technology at the heart of the platform. In this article we explain how it works, what you can realistically expect, and which habits help you get the most out of it.
Why Deduplication Matters for Proxmox Backup Server
Traditional backup tools treat every backup as a new copy of your data. Even when a virtual machine has changed just one log file since yesterday, a classic solution may still write another complete image to disk. Multiply that by dozens of VMs and a month of retention, and your storage costs explode. Proxmox Backup Server takes a fundamentally different approach: it looks at your data at a granular level and stores each unique piece exactly once — no matter which machine, date or snapshot it came from.
The result is a datastore that behaves almost like a living archive. Retaining daily backups for a year costs roughly the size of your environment plus the daily changes, not twelve months of full copies.
How Chunking Works Under the Hood
When a backup job runs, Proxmox Backup Server splits the incoming data stream into small pieces called chunks. The chunk boundaries are calculated using a rolling-hash algorithm, which means that inserting or deleting data at the beginning of a disk does not shift every boundary — only the affected region re-chunks. Each chunk is then:
- Hashed with a cryptographic fingerprint that identifies its content.
- Compressed with the ZSTD algorithm, which trades a little CPU for a lot of space.
- Stored in a content-addressable datastore, referenced by its hash.
Because the reference is based on content, two identical chunks — even from two completely different virtual machines — are stored once and referenced twice. An empty disk block that appears in a hundred VM images costs the space of a single chunk.
What Realistic Savings Look Like
Exact numbers depend on your workloads, but real-world reports are consistently dramatic. A typical Proxmox VE cluster with ten to thirty virtual machines usually sees the first full backup consume roughly the logical size of the data. After that, each nightly job only adds the chunks that actually changed — often just a few gigabytes even for large environments. Databases, mail servers and desktop images with lots of empty space compress and deduplicate extremely well; already-encrypted or compressed source data sees fewer gains, since its content looks random.
A good rule of thumb: after the first month, the datastore rarely grows faster than your genuine data churn — not the number of backups you keep.
Pruning and Keep Options: Keeping Growth Under Control
Deduplication alone does not manage retention — pruning does. Proxmox Backup Server offers simple keep rules such as keep-daily, keep-weekly and keep-monthly. When a snapshot no longer matches any rule, the server removes its references, and chunks that are no longer referenced by any snapshot are reclaimed during garbage collection. This three-part dance — deduplication, pruning and garbage collection — is what keeps a healthy datastore stable in size for years.
Getting Started with Your Own Datastore
None of this requires special hardware or a complicated setup. A modest machine with fast disks is enough, and after completing your proxmox backup download and installing the ISO, the first datastore can be created in the web interface within minutes. Point your Proxmox VE cluster at it, run the first backup, and watch the usage graph — most people are genuinely surprised by how little space their second and third backups actually consume.
