build a server →

Storage and backup servers.
Restores cost nothing.

Large arrays fed over a port with no counter. Metered hosting makes writing a backup cheap and reading it back expensive, which is the wrong way round.

port
1, 10, 25, 40, 100 Gbps
ingress
unmetered
egress
unmetered
locations
AMS · MEP · IAD · PHX
in plain words

Written every night, needed all at once.

A storage or backup server holds a lot of data and hands it back when asked. Backups are written regularly and read rarely, until the day you need all of them together.

That day is the one that counts. On a metered plan the restore is the expensive part, so people hesitate at the moment they should not. Here the port has no counter, which turns the question into how fast you need the data back, and the answer is the port size.

sizing it · four steps

How to choose.

  1. 01 ─ data

    Count what you must keep: today's size plus growth over the term you plan. Capacity comes first; the disk layout is agreed per build.

  2. 02 ─ window

    Set a restore window, the hours you can live without the data. Data size divided by that window gives the port speed.

  3. 03 ─ second copy

    A RAID array survives a failed disk, not a fire or a mistaken delete. Keep the off-site copy in another location.

  4. 04 ─ test

    A backup you have never restored is a guess. The calculator further down shows how long a full restore takes on each port.

what it solves
  1. 01 ─

    Restores cost more than backups

    on a meter or a shared host

    Metered hosting makes writing a backup cheap and reading it back expensive.

    here

    Ingress and egress both unmetered, so a full restore costs nothing extra.

  2. 02 ─

    RAID is not a backup

    on a meter or a shared host

    An array survives a failed disk, not a fire, ransomware or a mistaken delete.

    here

    Keep the off-site copy in another of our four locations, on the same unmetered port.

  3. 03 ─

    Plans sized for compute

    on a meter or a shared host

    General-purpose plans give you cores you do not need and disks you do.

    here

    Built around the array: capacity first, agreed with you on the build.

the right build for it

What to look for in the machine.

diskslarge arrays, sized to the data

Capacity comes first; the exact layout is agreed per build.

portsized to the restore window

How fast you need the data back decides the port, not the daily writes.

cpumodest

Parity and checksums, not compute. Enough cores for the array and little else.

The servers that fit this today are marked under the processor on the price list. The shelf changes; what to look for in the machine does not.

work it out · your numbers

How long a full restore takes.

Pick how much data comes back. Each row is the same restore on a different port, drawn to scale.

a full restore at line ratearithmetic · decimal TB
disks counted as
disks20 TB → 8 × 7.68 TB NVMe
1 Gbps1 d 20 h · stock
10 Gbps4 h 26 min · stock
25 Gbps1 h 46 min · custom build
40 Gbps1 h 6 min · custom build
100 Gbps26 min · custom build
The port sets the ceiling; the disk array and the protocol decide how close a restore gets to it. Reading the data back is not billed on any port: ingress and egress are both unmetered.
open this build in the configurator →

carries over: storage 8 × 7.68 TB NVMe

before you order

Questions to settle before ordering.

  1. 01 ─

    What happens if a disk in the array fails?

    We replace failed hardware at our cost (Terms, section 8). The data on it is yours to protect: we keep no snapshot and no copy of your server (section 9).

  2. 02 ─

    How are the disks arranged?

    Any type and size the job needs: SSD, NVMe or HDD. On two disks with Linux the default is a software mirror; with more disks, the RAID level you need, software or hardware. A hardware controller is fitted when the job needs one.

  3. 03 ─

    Do you sell backup storage as a service?

    Not here. HostiServer sells backup storage separately, as its own purchase; nothing in our terms gives you one by default (section 9).

  4. 04 ─

    What happens to the data when I cancel?

    Cancelling takes effect at the end of the billing period, and on termination the data on the server is deleted, so copy it off first. An unpaid invoice can lead to suspension after four days and termination seven days after that (sections 6 and 11).

The rest, for every machine: all questions and the terms.

where we draw the line

Tell us what it serves.

The workload, the working set, where the users are. We come back with a machine sized to it, a price and a date.

send the spec →compare servers →

alsounmetered servers · CDN origins · streaming servers · game servers · download servers · database & compute · virtualisation hosts · blockchain node servers