Skip to main content

AI Infrastructure

Persistent storage next to your GPUs

Two kinds of storage for EcoHash GPU instances and clusters: a Cloud Drive for one instance, a Shared Filesystem for many. Both survive the instance they were attached to and sit in the same region as the GPUs that read them.

Two storage types

One instance, or many

The difference is who can mount it at the same time.

Cloud Drive

Block storage (read-write-once) attached to one GPU instance at a time. A durable home directory for a single workspace: checkpoints, notebooks, rendered frames.

Attaches to GPU instances. Default mount /workspace/data-01.

Shared Filesystem

CephFS-backed shared storage (read-write-many) mounted by many instances at once and by every replica of a GPU cluster. Put datasets and model weights here once and read them from everywhere.

Attaches to GPU instances and clusters; also read by fine-tuning and dedicated inference. Default mount /workspace/shared-01.

How it works

Create once, attach as often as you like

Storage is a resource of its own, not a property of an instance.

  1. 1

    Create

    Pick a type and a size (50–500 GB) in the region where your GPUs run.

  2. 2

    Attach

    Select it when launching an instance or cluster — or create it inline in the same form.

  3. 3

    Work

    It appears under /workspace next to the 400 GiB local scratch that every GPU gets.

  4. 4

    Terminate

    Stop the instance whenever you like. The volume stays, with everything on it.

  5. 5

    Re-attach or export

    Mount it on the next instance, or export a folder to a download link.

What you get

Built for datasets and model weights

Outlives any instance

Storage is a separate resource. Terminate a GPU instance and the volume is ready to attach to the next one.

One copy of the weights

A Shared Filesystem is the same path on every replica of a cluster, so a 60 GB model is downloaded once, not once per GPU.

Fixed sizes, predictable bill

50, 100, 200, 300, 400 or 500 GB. Each size is a flat monthly rate; see the live table below.

Lives next to the GPUs

Volumes are region-local, so reads and writes never cross a region boundary.

Export any folder

Browse the volume from the console and turn a folder into a download link that stays valid for three days.

5-day grace period

If your balance runs out, suspended storage stays readable for five days so you can get your data out.

Pricing

Flat monthly rate per size

Rates are pulled live from the billing database; the pricing page carries the same table.

TypeSizePrice / month
Cloud Drive50 GB$5.00
Shared Filesystem50 GB$5.00
Cloud Drive100 GB$10.00
Shared Filesystem100 GB$10.00
Cloud Drive200 GB$20.00
Shared Filesystem200 GB$20.00
Cloud Drive300 GB$30.00
Shared Filesystem300 GB$30.00
Cloud Drive400 GB$40.00
Shared Filesystem400 GB$40.00
Cloud Drive500 GB$50.00
Shared Filesystem500 GB$50.00

Full GPU, model and storage rates on the pricing page.

At a glance

The essentials

Cloud Drive
Block · read-write-once
Shared Filesystem
CephFS · read-write-many
Sizes
50 – 500 GB
Billing
Per GB, monthly rate

Attach storage to your next GPU

Create a Cloud Drive or Shared Filesystem in the launch form and it mounts under /workspace automatically.