Selection path
First choose object, block, or file, then evaluate sharing scope, AZ boundary, performance model, lifecycle, encryption, and recovery objectives.
| Scenario | Prefer |
|---|---|
| Massive objects, static content, data lake | S3 |
| EC2 boot volume or database block storage | EBS |
| Shared POSIX files for Linux EC2 | EFS |
| Windows SMB, Lustre, NetApp ONTAP, OpenZFS | The matching Amazon FSx option |
| Hybrid files, volumes, or virtual tape | Storage Gateway |
| Large offline migration | Snow Family |
Frequent questions
- S3 strong consistency does not remove every need for retry and idempotency.
- Versioning, replication, Object Lock, and MFA Delete protect different failure paths; replication is not backup.
- EBS belongs to one AZ; a snapshot can create volumes in other AZs in the Region.
- EBS Multi-Attach requires a supported volume, one AZ, and cluster-aware software.
- An EFS mount target is an AZ network entry; clients normally need NFS TCP 2049.
- For private S3 delivery through CloudFront, prefer OAC over a public bucket.
Answer structure
Access protocol → consistency and performance → failure domain → encryption and authorization → lifecycle and cost → backup and DR.
Remember
S3 for objects, EBS for blocks, EFS for shared Linux files; discuss performance, availability, and backup separately.