Object storage keeps data as discrete objects, each with its own metadata and a unique key, in a flat namespace rather than a folder tree. It is the model behind services such as Amazon S3, and it is a common home for backups, media files and data lakes. This guide explains how it works, which features to look for, and how the major providers compare, including IBM Cloud Object Storage against AWS.
What is object cloud storage?
Object cloud storage saves each piece of data as an object. An object has three parts:
- Data: the contents, such as an image, a log file or a database dump.
- Metadata: descriptive information such as content type, custom tags and system details.
- A unique key: the name used to find the object.
Objects live in containers called buckets. Azure calls them containers. Bucket names must be unique, and in Amazon S3 they are unique across all AWS accounts. Amazon S3 and Google Cloud Storage both use this model.
Applications typically reach objects over HTTP, using operations such as PUT (upload), GET (download), HEAD (read metadata), DELETE and LIST. The S3 API has become the de facto standard, and many other providers implement it.
How object storage is organized
Object storage has a flat namespace, so there are no real directories. A “folder” in the console is a shared key prefix. The key photos/2026/trip.jpg is one string. The console shows a folder-like view by listing keys that start with photos/2026/.
This has practical effects. Renaming a “folder” that holds millions of objects means copying and deleting every object under that prefix, which can take a long time. It is not a quick metadata change.
Objects are written whole. To change an object, you upload a new version of it. Some services support limited appends, but you generally cannot edit a few bytes in the middle of a file the way you can on a disk.
Key features of object storage
- Durability through replication or erasure coding. Providers spread copies or encoded fragments across devices and sites. Amazon S3 Standard is designed for 99.999999999% durability (11 nines) and stores data across at least three Availability Zones.
- Versioning. Keeps earlier versions of an object, which helps you recover from accidental overwrites and deletes.
- Lifecycle rules. Move objects to cheaper storage classes, or delete them, after a set time.
- Object lock and retention. Prevent objects from being changed or deleted for a set period. This helps against ransomware. Amazon S3 Object Lock and Google Cloud retention policies are examples.
- Replication. Copies objects to another region or bucket. Cross-region replication typically has to be configured.
- Event notifications. Trigger a function or workflow when an object is created or deleted.
- Presigned URLs. Grant time-limited access to one object without sharing credentials. With AWS Signature Version 4, the maximum lifetime is 7 days.
- Custom metadata. Attach tags and fields to objects so they can be searched and classified.
Object storage providers compared
| Provider | Service | Notes |
|---|---|---|
| AWS | Amazon S3 | The widest range of storage classes, regions and integrations |
| Microsoft Azure | Azure Blob Storage | Hot, Cool, Cold and Archive tiers; containers group blobs within a storage account |
| Google Cloud | Google Cloud Storage | Standard, Nearline, Coldline and Archive classes; Autoclass can move objects between classes |
| IBM Cloud | IBM Cloud Object Storage | S3-compatible API; built on Cleversafe technology |
| Cloudflare | R2 | S3-compatible API; marketed with no egress fees, so check current terms |
Self-hosted options include MinIO and Ceph RADOS Gateway, which provide S3-compatible APIs on your own hardware.
IBM Cloud Object Storage vs AWS
IBM Cloud Object Storage (COS) traces its design to the Cleversafe platform, which IBM acquired in 2015. It spreads data across sites using information dispersal, a form of erasure coding, and it exposes an S3-compatible API, so many S3 tools can work with it.
When comparing IBM COS with AWS S3, keep these points in mind:
- Feature breadth. AWS S3 offers more storage classes, regions and integrations with other AWS services. IBM highlights features such as Smart Tier, which moves data between tiers automatically, and its high-speed transfer tools.
- API compatibility. S3-compatible does not guarantee full feature parity. Before you commit, check the specific features you need, such as object lock, lifecycle rules, event notifications and access policies, against the provider’s documentation.
- Price. Costs depend on region, storage class and the volume of requests and data transfer. Compare current price lists using your expected usage, not headline per-gigabyte rates.
Do you have to pay for object storage?
Usually, yes, once you store more than a small amount. Consumer services offer a free allowance. Cloud providers bill for the data you store, the requests you make, and data transferred out. Some providers include a small free tier. Those free tiers are for testing, not for production data volumes.
Common misconceptions
- Object storage is not a file system. You can’t mount it like a normal disk. Tools such as s3fs and Mountpoint for Amazon S3 emulate a file system, with limits.
- Console folders are not real directories. They are key prefixes.
- High per-request latency does not mean low throughput. Applications making many parallel requests can reach high aggregate throughput.
- Durability and availability are different. Durability describes the chance of losing data. Availability describes whether you can reach it right now.
- “Unlimited” does not mean free. Storage, requests, retrieval and data transfer out all cost money.
When object storage is a good fit
- Backups and archives
- Images, video and other media files
- Logs and analytics data in a data lake
- Static website assets (Amazon S3 can serve static sites over HTTP, and HTTPS is usually added through a CDN such as CloudFront)
- Machine learning datasets
For a side-by-side look at object and block storage, read why use object storage instead of block storage?. For the wider picture, see types of cloud storage services. If you want to know how applications connect to S3, see Is S3 a protocol or API?