Storage: S3, EBS, EFS & Friends
Object vs block vs file storage, S3 storage classes, and which storage service each scenario calls for.
Three kinds of storage
| Type | Think of it as | AWS service | Attach to |
|---|---|---|---|
| Object storage | An infinite bucket of files with URLs | Amazon S3 | Accessed over the network by anything |
| Block storage | A virtual hard drive | Amazon EBS | One EC2 instance at a time (same AZ) |
| File storage | A shared network drive | Amazon EFS (Linux) / FSx (Windows, Lustre) | Many instances simultaneously |
Amazon S3 essentials
Key points
- Stores objects in buckets; virtually unlimited capacity; objects up to 5 TB.
- 11 nines (99.999999999%) durability — data is redundantly stored across multiple AZs.
- Common uses: backups, data lakes, static website hosting, media storage.
- Versioning protects against accidental deletion; lifecycle policies move objects between storage classes automatically.
- Access controlled by bucket policies, IAM, and Block Public Access (on by default).
S3 storage classes
| Class | For | Trade-off |
|---|---|---|
| S3 Standard | Frequently accessed data | Highest cost, instant access |
| S3 Intelligent-Tiering | Unknown or changing access patterns | Auto-moves objects between tiers for you |
| S3 Standard-IA / One Zone-IA | Infrequently accessed but needs fast retrieval | Cheaper storage, per-GB retrieval fee (One Zone = single AZ, less resilient) |
| S3 Glacier Instant / Flexible Retrieval | Archives (instant, or minutes-to-hours retrieval) | Very cheap storage, retrieval costs/time |
| S3 Glacier Deep Archive | Long-term archives (7–10 yr compliance) | Cheapest of all; retrieval takes ~12+ hours |
Storage-class questions hinge on two words: access frequency and retrieval time. "Rarely accessed, must retrieve immediately" → Standard-IA. "Compliance archive, retrieval can wait hours" → Glacier Deep Archive. "Unpredictable access" → Intelligent-Tiering.
Block and file storage
Persistent virtual disks for EC2. Snapshots back them up to S3. An EBS volume lives in one AZ.
Physically attached, ultra-fast ephemeral disk — data is lost when the instance stops.
Elastic NFS file system shared by many Linux instances across multiple AZs; grows and shrinks automatically.
Managed file systems: FSx for Windows File Server (SMB/Active Directory) and FSx for Lustre (high-performance computing).
Hybrid bridge letting on-premises apps use S3-backed cloud storage.
Central, policy-based backups across EBS, RDS, DynamoDB, EFS, and more.
Keyword map: shared Linux file system → EFS; Windows file share → FSx for Windows; boot volume / database disk for one instance → EBS; temporary scratch space → instance store; on-prem apps using cloud storage → Storage Gateway.
Which AWS service provides object storage with 99.999999999% (11 nines) durability?