Storage server rental: how to work out usable capacity and the price per terabyte
This selection is built for backups, archives, video and file storage: physical servers with at least 4 TB of hard disk space in total. Plans are sorted by price per terabyte, not by the price of the server. As of 3 November 2026 there are 746 plans from 41 hosts. The cheapest server costs about $21 a month, the median is about $260, and the median price per terabyte is about $36 a month. The most common size is 4 TB of disks, followed by 8, 12 and 16 TB, and there are servers with 24 and 48 TB. The processor matters less here: mostly Xeon E5-2620 v3 and v4, dual Xeon E5 and Silver, Core i9-9900K and Ryzen 7 3700X. Most offers are in Germany, then the US, Russia, France and the UK.
Raw capacity and usable capacity
A plan lists the total size of all its disks. How much of that is left for data depends on how the disks are arranged.
- RAID 1 (mirror). Two 4 TB disks give 4 TB of usable space. Survives the loss of one disk.
- RAID 5 or RAIDZ1. One disk goes to parity: four 4 TB disks give 12 TB. Survives the loss of one disk.
- RAID 6 or RAIDZ2. Two disks go to parity: four 8 TB disks give 16 TB. Survives two disks failing at the same time.
- RAID 10. Half the capacity, but faster writes and faster rebuilds.
So compare the price of a usable terabyte, not a raw one. A 2×8 TB server in a mirror gives the same space as 4×4 TB in RAID 5, but the two usually cost different amounts.
Why RAID 5 on large disks is risky
When a disk fails, the array rebuilds: the remaining disks are read in full and the data is recalculated onto the new one. On 8–16 TB disks this takes anywhere from a day to several days, and the array runs without protection the whole time. If a second disk fails or an unreadable sector turns up during that window, data is lost. For disks of 8 TB and up, RAID 6 or RAIDZ2 is the safer choice.
Another trap is shingled (SMR) disks. They are cheaper, but rewrite data very slowly, and during a rebuild, especially under ZFS, they can drop out of the array on timeouts. Manufacturers label disks meant for arrays as CMR. Ask the host for the disk model and check it against the manufacturer's spec sheet.
Network and traffic matter more than the CPU
A storage server has to be filled and read over the network. Do the maths: 8 TB over a 1 Gbit/s port at full load takes almost 18 hours. If you plan an initial archive upload or regular restores of large volumes, this matters.
Check the traffic allowance too: some plans in the selection pair 8–16 TB of disks with only 5 TB of traffic a month. That is enough for an archive that is only written to, but not for serving video. Plans without traffic metering are in the unlimited traffic servers selection, fast ports in the 10 Gbit/s servers selection.
What these servers are used for
- Backups of other servers. A machine in another data centre that receives copies of databases and files every night. Capacity and reliability matter most; speed only counts during a restore.
- Video and photo archives. CCTV recordings, production studio source files, a stock photo library. Volumes grow by terabytes a month, so find out in advance whether disks can be added.
- Team file storage. Shared folders, file sync for staff, S3-compatible storage for applications.
- Serving static content. If many users download the files, port speed and traffic become requirements alongside the disks.
What to check before renting
- Disk model and condition. Once the server is handed over, run
smartctl -aon every disk: power-on hours, reallocated and pending sectors. Ask for disks with high hours and errors to be replaced straight away. - Memory for ZFS. ZFS uses memory as a read cache, and the Proxmox documentation suggests about 1 GB of RAM per terabyte of storage. With 32 TB of disks, 16 GB of RAM is already tight.
- Disk replacement time. Ask how many hours it takes to replace a failed disk and whether the server has to be powered off.
- Adding disks later. If the chassis has free bays, storage can be expanded without moving.
Most importantly, RAID is not a backup. It protects against a disk failing, not against deleted files, a software bug or a problem with the server itself. If this storage is your backup, keep a second copy of the most important data in another data centre. For smaller volumes, see VPS for backups and VPS with a large disk.