---
title: "Drive"
description: "Disks of a project that outlive the sandboxes and Cubes that use them. One sandbox writes a Drive at a time; it is copied to its region's object store as it is written, so any machine of the region can mount it."
---

> Documentation Index
> Fetch the complete documentation index at: https://glac.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Drive

A Drive is a disk of your project that lives on after the sandboxes that use it. Inside a sandbox it is a block device with its own ext4 file system, mounted at a path you choose, and its blocks are on the machine's own NVMe disk. One sandbox has it at a time, read-write. When that sandbox goes, the Drive keeps what it holds, for the next sandbox.

As a Drive is written, it is copied into its region's object store. So a sandbox on any machine of the region can mount it: if another machine held it last, it is brought over first, block by block as they are read. A sandbox of another region mounts it too, where that region's machines read this one's object store: its blocks come from its home region as they are read, and its home follows it there after an hour ([across regions](/docs/drive/concepts/#across-regions)).

## When to use Drive

- **A database or a queue**: a [Cube with a Drive](/docs/cube/services/#a-service-that-keeps-data) has one instance, which keeps its data across deploys and restarts.
- **An agent's workspace** that outlives its sandbox: mount the same Drive in the next sandbox.
- **One disk per user or session**: each [keyed instance](/docs/cube/keyed-instances/#a-drive-for-each-key) of a Cube gets a Drive of its own.
- **Large working data** that a sandbox's own disk should not hold.
- **One data set for many readers**: [mount a snapshot](/docs/drive/snapshots/#mounting-a-snapshot-read-only) of a Drive read-only in any number of sandboxes at once.
- **A copy to try something on**: [fork](/docs/drive/snapshots/#forking-a-drive) a Drive, and delete the fork when you are done.

For a file system that several sandboxes mount at once, see [shared volumes](/docs/drive/volumes/).

## How it works

1. **Make.** `POST /v1/drives` (`glac drives create`) with a name, a size and a region. Only what is written takes space and is billed.
2. **Mount.** Give it to a new sandbox (`drives` when you make it), to a running one (`POST /v1/sandboxes/{id}/drives`), or to a Cube's version.
3. **Keep.** Unmount it, or delete the sandbox: the Drive stays with what it holds, in its region, until you delete it.
4. **Snapshot.** `POST /v1/drives/{id}/snapshots` (`glac drives snapshots create data`) keeps the Drive as it is now, without copying it. Restore the Drive to a snapshot, fork a new Drive from it, or mount it read-only: see [snapshots and forks](/docs/drive/snapshots/).

## What it costs

| Meter | Price | Free, each day | Paid includes a month |
|---|---|---|---|
| A Drive’s data, by the bytes written, all of it in tiered storage | $0.02 a GB-month | 15 GB of data in all, with sleeping instances and snapshots | 20 GB-months of data, with sleeping instances and snapshots |
| A Drive’s NVMe cache, the size set | $0.06 a GB-month | 15 GB in all | 20 GB-months |
| Writing and reading a Drive | No charge |  |  |

A Drive is billed by its data and its NVMe cache. Its data is counted by the bytes written to it and not trimmed away, not by its size, hour by hour, all of it kept in tiered storage; its cache is the part of it kept on the machine's NVMe, the size you set: all of it for local speed, none for the lowest price, or a balance between. Writing and reading cost nothing. Free's data is shared with sleeping sandboxes and snapshots. A snapshot of a Drive is billed for what it keeps alone, at the same price: see [snapshots and forks](/docs/drive/snapshots/#what-is-billed) and [pricing and plans](/docs/platform/pricing/).

## Limits of each plan

|  | Free | Paid |
|---|---|---|
| Drives and volumes of one project | 3 | 500 |
| Largest Drive | 10 GiB | 1 TiB |
| All Drives of a project together | 15 GiB | No limit |
| Drives and volumes one sandbox mounts | 8 | 8 |

The number of Drives of a paid plan is a soft limit: ask us to raise it. An enterprise account can be given Drives of up to 16 TiB. See the [limits page](/docs/platform/limits/).

## Not there yet

- A Cube's version mounts its Drives read-write only: read-only mounts of a snapshot are for sandboxes.

## Next

- [Quickstart](/docs/drive/quickstart/): make a Drive, write to it from one sandbox, read it from the next.
- [How Drives work](/docs/drive/concepts/): mounting, one writer, where a Drive is, what is billed.
- [Snapshots and forks](/docs/drive/snapshots/): keep a Drive as it is, restore it, fork it, mount it read-only in many sandboxes.
- [Shared volumes and backups](/docs/drive/volumes/): file systems several sandboxes mount at once.

Source: https://glac.ai/docs/drive/index.mdx