Object storage

Also known as blob storage, object storage is a technology that stores and manages data as discrete units called objects. Rather than organising data into a hierarchy of files and folders (as file storage does), object storage uses a flat namespace: all objects live in a single container called a bucket, each identified by a unique key. Objects can be anything — images, videos, audio files, backups, log files, machine learning datasets.

Each object consists of three things:

  • Data: the raw binary content of the file.

  • Metadata: a flexible, arbitrarily rich set of key-value pairs describing the object (content type, creation date, author, tags, and any custom application-specific attributes).

  • Unique identifier: a globally unique key used to retrieve the object via an HTTP/REST API.

Unlike file storage (NFS/SMB) or block storage, object storage is not mounted as a filesystem. Data is accessed programmatically via an API (the de facto standard being Amazon S3’s API). This design removes the need for a traditional file-system layer, which is what allows object storage to scale to virtually unlimited capacity across multiple physical devices, data centres, and regions.

Advantages

  • Massive scalability: No hierarchical namespace to become unwieldy at scale. Systems like Amazon S3 routinely store trillions of objects and exabytes of data.

  • Rich metadata: Custom metadata per object enables sophisticated search, organisation, and lifecycle policies without a separate database.

  • Durability and availability: Objects are automatically replicated across multiple devices and availability zones. Amazon S3, for example, is designed for 99.999999999% (eleven nines) durability.

  • Cost-effectiveness: Per-gigabyte costs are typically far lower than block or file storage, making it well-suited to large-volume, infrequently-modified data (backups, archives, media).

Limitations

  • Higher latency: HTTP-based retrieval is slower than block-level I/O. Object storage is unsuitable for workloads that need sub-millisecond latency (databases, VMs).

  • No partial updates: An object must be replaced in full to update any part of it. Frequently-modified data is better served by block storage.

  • No filesystem semantics: There is no concept of directories, file locking, or POSIX permissions. Applications must interact via the API.

Best for

Object storage is the storage of the cloud, best suited to large volumes of unstructured data that is written once and read many times. Common use cases include:

  • Static assets for web applications (images, CSS, JavaScript).

  • Media storage and content delivery (videos, audio).

  • Data lakes and machine learning training datasets.

  • Application backups and disaster recovery archives.

  • Log file aggregation.

Cloud implementations include Amazon S3, Azure Blob Storage, and Google Cloud Storage.

See Also