Skip to main content

Create a storage device

how-to · Reviewed 2026-07-31 · FEAT-0118 GA

Two choices here are hard to undo. The coretype caps how much you can ever extend the device, and walltime cannot be pushed past 365 days from the cluster's start.

Create it​

Storage → New Storage Device. The form has three groups (Storage Settings, Hardware, and Alerts) plus a live Hourly Price Summary in the left rail.

FieldNotes
Storage SizeDecimal GB/TB. The OS and filesystem consume some of it
Auto-extendAt a set percent full, the disk grows automatically
Coretype and CoresUnder Hardware. Caps how far you can extend later — see below
RegionMust match where your jobs run
WalltimeDays before automatic termination. Max 365 days
First / Second AlertDisk-full percentages that email you
The Hourly Price Summary updates as you size the device — an HPS bills like a running job, not like object storage.

The coretype list shows IOPS per core for each option, which is the number to compare if the device will be doing real I/O.

Provisioning takes up to 15 minutes. Click the device in the list to see details and browse files.

Project budgets can terminate the device

A device tied to a project draws on that project's budget. If a strict budget runs out, the device is terminated: see budgets and deposits.

Choose the coretype carefully​

Pick a bigger coretype than you think you need

How much you can extend, and how many times, is capped by the coretype and core count you choose here, and the UI doesn't show you that limit. You find out when an extension request fails.

If you expect to grow the device, choose a more capable coretype and more cores up front. A larger instance supports more disks, more capacity, and more extensions. Under-specifying is not recoverable without rebuilding the device.

The coretype also sets the ceiling on attached jobs, so under-specifying constrains you twice.

Sizing and performance​

AWSAzure
Maximum size256 TiBDepends on coretype, core count, and starting size
Extension frequencyMay enforce a 6-hour gapEvery few minutes
Extension durationMinutes, but up to 24 hours for a large deviceConsistently minutes

Throughput and IOPS scale with disk size. A bigger disk is a faster disk. For moving large volumes (100 GB and up), choose more cores too; transfer throughput is CPU-bound as well as disk-bound.

There's also a minimum size per coretype: each instance type ships with bundled storage, and that bundle is the floor for that coretype and core-count combination. An error about a minimum storage size means you picked below it.

Alerts and auto-extend​

Set both. They do different jobs: alerts tell you, auto-extend acts.

Auto-extend multiplies capacity, and can fire repeatedly. On AWS the filesystem doubles; on Azure it grows by a larger multiple. The multiplier is fixed and you can't choose it.

Auto-extend can multiply your bill without approval

HPS is billed by size. A runaway job that fills the disk triggers a doubling, then potentially another. Nobody signs off on the increase.

Set the alert thresholds even if you enable auto-extend, so you find out while it's happening.

Alerts and auto-extend run on a periodic check

Neither is instantaneous. A rapidly filling disk can pass the threshold before either fires.

Add files​

Once the device is running, from its file browser:

  • Create Folder — the new folder lands under the selected directory.
  • Use files from Cloud — pull from Rescale Files.
  • Upload from this computer — the file goes to Rescale Files first, then copies to the device.
  • Add Job — pick one or more jobs; their inputs and outputs land under the jobs directory.
You can't download straight from a device

Data must sync to Rescale Files before you can download it. For direct access, use SFTP.