Replicated network NVMe that answers in ~160 microseconds: local-disk speed with none of the local-disk risk. Four classes, hot-attach, resize and retier in place, first 25 GB free on every VM.
A local disk ties your data to one machine: when that machine dies, the data dies with it. Our fabric closed the latency gap, so you keep local-NVMe speed and lose the single point of failure.
Every write lands on independent storage nodes before it is confirmed. A dead disk or a dead host changes nothing for your data.
Data moves over a direct memory path between machines, skipping the kernel network stack. The number is from fio inside real guests, and you can reproduce it on yours.
Disks follow VMs through live migration and node failures. Volumes hot-attach and detach without reboots, and move between VMs freely.
One fabric, four personalities. Pick by workload; speed tiers change in place with no data migration.
The default
Replicated fast NVMe for almost everything: sites, apps, dev, everyday services. Five speed tiers from 5k to 200k IOPS.
The first 25 GB on a VM root disk are free.
From €0.038 /GB per month
Unlimited fsyncs
Enterprise NVMe with power-loss protection: an acknowledged write survives a power cut, and commits run at full device speed.
For Postgres, MySQL, queues, anything transactional.
From €0.058 /GB per month
Maximum speed
Single-copy NVMe with no meaningful throttle: the ceiling is your VM's hardware, measured at 203k mixed IOPS from an ordinary guest.
For caches, scratch and CI, where data is disposable and speed is everything.
€100 /TB per month
Terabytes, cheap
Replicated bulk space for archives, media and cold data. HDD-like speed profile at 200 MB/s, NVMe fabric and full replication underneath.
Volumes up to 10 TB.
€30 /TB per month
Limits here are delivery promises, not marketing ceilings.
A 5,000 IOPS tier delivered 5,016 IOPS in testing; a 1 GB/s tier held 1027 MiB/s flat. You get the number on the card.
Basic VMs cap at a real-world 50k IOPS because that is what the guest silicon can drive. High Frequency guests run the full tiers, verified past 300k.
An independent vpsbenchmarks.com run measured ~212k mixed IOPS on a 2 vCPU VM, roughly double the same-size plans in their charts, on a replicated disk.
Full engineering story, latency methodology and every fio command: Block Storage docs. Head-to-head against the local NVMe of Hetzner and DigitalOcean: the benchmark.
Replication protects you from our hardware. Snapshots and backups protect you from everything else.
Point-in-time copies on the cluster for fast rollbacks after a bad migration or a wrong command. Billed on compressed data only.
Encrypted with your organization's key before leaving the hypervisor, stored at third-party providers in other countries, recoverable without ServersCamp.
How it works: Snapshots & Backups docs · What it costs: Backups pricing
A production Postgres disk costs under €10 a month. A replicated terabyte, €30.
