Storage
Amazon S3
Durable object storage with tiered pricing
Amazon S3 is object storage built for 99.999999999% (11 nines) durability. Objects live in buckets within a Region and are replicated across multiple Availability Zones. Storage classes (Standard, Intelligent-Tiering, Standard-IA, One Zone-IA, Glacier tiers, Deep Archive) trade cost against access frequency and retrieval time.
When to use it
- Static assets, media, backups, and data lakes
- Content served through CloudFront
- Archival with lifecycle transitions to Glacier tiers
- Any workload needing virtually unlimited durable object storage
When NOT to use it
- Block storage for a running OS or database (use EBS)
- A POSIX file system mounted by many instances (use EFS or FSx)
- Low-latency sub-millisecond access
Key exam points
- Standard is multi-AZ and millisecond-latency; Deep Archive is cheapest with hours-long retrieval
- Lifecycle rules transition objects automatically by age
- Versioning protects against overwrite and deletion; MFA delete adds extra protection
- Gateway VPC endpoints reach S3 privately and avoid NAT data-processing fees
- Intelligent-Tiering auto-moves objects for unknown/changing access patterns
Compared with
- Amazon EBS — EBS is single-AZ block storage attached to one instance; S3 is multi-AZ object storage accessed over HTTP.
- Amazon EFS — EFS is a mounted file system shared by many instances; S3 is object storage accessed via API.