Amazon Simple Storage Service

S3

You need to store files that any number of machines can read and write, and you never want to think about disks running out of space.

Reach for it when

Storing user uploads, profile pictures, and PDF invoices for a web application.
Hosting static website assets such as HTML, CSS, JavaScript, and images directly.
Serving as a data lake for large-scale analytics datasets from Athena or EMR.

Storing frequently updated transactional application data with low latency — use RDS or DynamoDB instead.
Mounting a shared filesystem to legacy applications via standard POSIX protocols — use EFS instead.
Managing structured application logs that need sub-second query and search capabilities — use CloudWatch Logs instead.

Service	Pick it instead when
EFS	Choose it when you need a POSIX-compliant file system that can be mounted simultaneously by multiple EC2 instances.
EBS	Choose it when you need block storage with sub-millisecond latency mounted to a single EC2 instance.

The model: Pay per GB stored per month, plus fees for API request volume and external data transfer out.
The line item that surprises people: Enabling bucket versioning without lifecycle rules keeps every historic file version and deleted file billable forever.

A bucket policy granting s3:GetObject on arn:aws:s3:::bucket grants nothing on objects; object paths need the /* suffix.
The S3 consistent read model is immediate for new uploads but changes to existing keys can suffer from propagation delays.
Bucket names are globally shared across all AWS customers; once a bucket name is taken, you cannot use it in any region.

This page is a concept reference. Cost models are qualitative; confirm the current offering, Region and pricing before deploying.