Object Storage
Subject: Computer and Technology noun /ˈɑːb.dʒɪkt ˌstɔːr.ɪdʒ/ 1 definition
- storage-architecture
- cloud-storage
- distributed-systems
- data-storage
-
Definition (noun)
Entry #318 · UpdatedObject Storage is a computer data storage architecture that manages information as discrete, self-contained units—called objects—within a flat address space rather than a hierarchical directory tree or fixed-size disk blocks. Each object bundles an arbitrary binary or text payload with extensible key-value metadata and a globally unique identifier (key) inside a logical container typically called a bucket. Applications create, read, and delete objects over stateless RESTful HTTP APIs—most commonly following the Amazon S3 interface—while the underlying cluster distributes replicas or erasure-coded shards across independent nodes to scale horizontally to exabyte capacities. Unlike block devices or POSIX-compliant file systems, object storage treats each object as an indivisible unit without in-place byte modifications or native filesystem locking, trading sub-millisecond random I/O for massive throughput, custom metadata indexing, and high durability.
Common Types
Cloud-Hosted Object Storage: Managed, multi-tenant services—such as Amazon S3, Google Cloud Storage, and Azure Blob Storage—that automatically handle hardware provisioning, cross-zone replication, and lifecycle transitions behind pay-as-you-go HTTPS APIs.
Self-Hosted Software-Defined Object Storage: Distributed storage platforms like MinIO, Ceph RADOS Gateway (RGW), and OpenStack Swift deployed on private clusters or Kubernetes to expose S3-compatible APIs under full control of data residency and egress costs.
Content-Addressable Object Storage: Implementations that derive an object's key from a cryptographic hash of its payload rather than a user-assigned path, providing built-in deduplication and tamper detection in systems like Git, OCI container registries, and backup vaults.
Archival and Cold-Tier Object Storage: High-density, low-cost storage classes—such as AWS S3 Glacier Deep Archive or Google Cloud Archive—that require an asynchronous restore step before reads in exchange for minimal per-gigabyte pricing over multi-year retention periods.
Example
When a user uploads a 45 MB video through a media service, the backend issues an HTTP
PUTrequest tos3://media-prod/uploads/9f8c2a.mp4with custom metadata headers (x-amz-meta-owner-id: usr_4812) andContent-Type: video/mp4. The object store computes an ETag checksum, distributes erasure-coded fragments across multiple availability zones, and returns200 OKso transcoding workers can immediately fetch the object via a presignedGETURL without mounting a shared filesystem.