Block vs Object vs File Storage

The three storage types

	Block	Object	File
Mental model	Raw disk (hard drive)	Key → value bucket	Shared network folder
AWS service	EBS	S3	EFS
Access unit	Blocks (512B-4KB)	Whole objects	Files
Structure	Formatted FS on blocks	Flat namespace (keys)	Real directories
Writers	1 EC2	Many (API)	Many (NFS)
Price/GB	$0.08 (gp3)	$0.023	$0.30 (most expensive!)
Latency	Lowest	Medium	Medium

Block (EBS) — the hard drive

Attached to ONE EC2 (like internal SSD). Format it (ext4/NTFS), mount, use.
Filesystem tracks which blocks belong to which file (pointer table).
Lowest latency — OS disks, databases (high IOPS).
Snapshots → S3 for backup. Can’t share across EC2 (multi-attach rare/limited).

Object (S3) — the cloud bucket

Key → data + metadata + ID. Flat namespace, no real folders (prefixes only).
NOT a filesystem — no mount, no locking. HTTPS REST API (GET/PUT by key).
Unlimited scale, 11 nines durability (99.999999999%).
Multi-writer safe. Great for static files, media, backups, data lakes, static sites.

File (EFS) — shared network drive

POSIX: real dirs, permissions, locks. Accessed via NFS.
Attachable to MANY EC2 simultaneously — all see same files.
Elastic (grows/shrinks). Linux only.
Use: shared assets across web farm, home dirs, lift-and-shift POSIX apps.
FSx = managed file systems: Windows (SMB) / Lustre (HPC parallel).

Key exam distinctions

EBS = block, single EC2, snapshot to S3
S3 = object, 11 nines, unlimited, API-based
EFS = file/NFS, multi-EC2, Linux only
“formatted disk attached to one server” → EBS
“shared files across servers” → EFS
“store unlimited files via API” → S3

[[aws/index]]
[[aws/compute-map/index]]