Cloud
Object Storage
Stores files as objects — data plus metadata — in a flat namespace of buckets, addressed by key over HTTP, with effectively unlimited scale and very high durability (S3, GCS). It is perfect for images, backups, logs and static sites. It is not a filesystem, though: no in-place edits or low-latency random writes, so do not treat it like a disk.
Purpose
Object storage keeps files as immutable objects — data plus metadata, addressed by key — in flat buckets accessed over HTTP, with effectively unlimited capacity and famously high durability (S3's design target is eleven nines). It exists to make 'store this file forever, retrievable from anywhere' a solved problem.
When to Use It
User uploads and media, backups and archives, logs and data-lake storage, and static website hosting behind a CDN. Lifecycle policies move aging data to cheaper tiers (infrequent access, archival) automatically.
Trade-offs
It is not a filesystem: objects are replaced whole, not edited in place; listing is by key prefix, not directories; and per-request latency suits bulk reads, not random small writes. Egress bandwidth is where the bill quietly grows.
Implementation
Design keys around access patterns (prefixes as pseudo-folders), block public access by default, and grant access via pre-signed URLs so clients upload/download directly without proxying through your servers. Enable versioning for the buckets that matter, encrypt at rest, and set lifecycle rules for cost.