Guides

Storage options across providers: block, object and backup storage

The three storage categories every cloud offers, what each is for, how they are billed, and how to combine them, with Linode and Alibaba Cloud as examples.

Updated September 25, 2026

Storage options across providers: block, object and backup storage

Every cloud provider sells storage in the same three shapes, even when the product names differ. Block storage behaves like a disk. Object storage behaves like a bucket of files reached over an API. Backup storage is a copy of something else, kept somewhere safe. Choosing well means knowing which shape fits each kind of data and how each is billed. This guide uses Linode (Akamai Cloud) and Alibaba Cloud as examples because they name the categories differently while offering the same underlying model.

Block storage: a disk you attach

Block storage is a volume attached to a single server and formatted with a filesystem. The operating system sees it as a drive. It is the right home for anything that needs low-latency random access: databases, application working directories, container image caches. On Linode it is called Block Storage; on Alibaba Cloud, disks attach to Elastic Compute Service instances and can be resized independently.

Two properties matter. A volume is attached to one instance at a time, so it is not a way to share files between servers. And it is billed by provisioned capacity, so a mostly empty volume costs the same as a full one. Size volumes to what you need plus growth headroom, not to what you might need in three years.

Local versus attached

Many instance plans include local storage on the host. It is fast and included in the instance price, but it usually disappears when the instance is deleted. Attached block volumes persist independently, which is why the database goes on a volume and the operating system can live on local storage.

Object storage: files behind an API

Object storage holds files as objects in buckets, reached over HTTP with an S3-compatible API in most cases. It scales without provisioning, can be reached from many servers or from the internet at once, and is billed by what is stored plus what is transferred. Linode Object Storage and Alibaba Cloud Object Storage Service both fit this description.

It is the correct place for user uploads, static assets, exported reports, logs and, critically, backups. It is the wrong place for a database file or anything that needs to be opened and modified in place.

Question Block storage Object storage
How is it accessed As a disk on one server Over an API from anywhere
What is it for Databases, working data Uploads, assets, backups, archives
How is it billed Provisioned capacity Stored capacity plus transfer
Can it be shared No Yes

Backup storage: the copy that saves you

Backup storage is not a separate technology so much as a separate purpose. Provider snapshots capture a whole volume or instance and are the fastest way to roll back. Application-level backups, such as database dumps written to object storage, restore across instance sizes and providers. Both belong in a baseline; the backup and recovery article describes how to layer them and why one copy should live outside the primary account.

Snapshot billing is usually by stored size, and snapshots accumulate quietly. A retention policy, enforced by automation, keeps the line on the invoice from growing every month; the first invoice guide lists snapshots among the usual surprises.

A pattern that works on any provider

  1. Operating system on the instance’s included storage.
  2. Database and application state on an attached block volume sized with headroom.
  3. User uploads and static assets in object storage, served directly or through a CDN.
  4. Nightly database dumps to object storage, encrypted.
  5. Scheduled snapshots of the block volume with a fixed retention.
  6. A copy of the dumps in a second account or provider.

This pattern separates data by how it is accessed and how it must survive, which is what keeps both performance and cost predictable.

Provider notes

On Linode, Block Storage, Object Storage and instance backups are all available from one account, and the Linode $100 Credit Account listed on the marketplace is enough to try the full pattern at small scale. Alibaba Cloud pairs Elastic Compute Service disks with Object Storage Service and snapshots across a large set of regions, which matters when uploads must stay near users; the Buy Alibaba Cloud Account page lists two Alibaba Personal Account options. Confirm regional availability and any credit coverage for storage services before planning around either. For choosing the compute that sits beside the storage, see choosing instance sizes.

Questions on this topic

Can I use object storage as a database disk?

No. Object storage is accessed over an API as whole objects, not as a block device with random read and write access, so databases cannot run on it. Put the database on an attached block volume or local NVMe and use object storage for the database dumps. The two categories complement each other rather than substitute for one another.

How is cloud storage typically billed?

Block storage by provisioned capacity, whether or not it is full. Object storage by stored capacity plus data transferred out, and sometimes by request count. Snapshots by the size retained. Exact rates vary by provider and region, so read the pricing page for the specific service and set a retention policy on snapshots to keep that line from growing.

Which storage should hold user uploads?

Object storage. It scales without resizing, can be reached from every application server at once, and can serve files directly or through a CDN. Keeping uploads off the application server's disk also makes servers disposable, which simplifies scaling and recovery. See the Linode FAQ for questions about the storage services on that platform.

More from the Blog

All articles