Storage Fundamentals¶

Notes from AWS Apprenticeship — February 2026.

Ultra-Short Summary¶

Storage underpins everything in cloud — your code, logs, data, backups, secrets. Understanding what can go wrong (and how AWS prevents it) gives you the mental model for why services like S3 versioning, EBS snapshots, and RDS automated backups exist. The GitLab and Toy Story 2 incidents are the canonical examples: even well-designed systems lose data when backup processes aren't validated.

Why Storage Matters More Than You Think¶

The GitLab Incident (2017)¶

Production engineer ran rm -rf on the production database
GitLab had 5 backup mechanisms — none worked correctly
Root cause: backups configured but never tested
Lost 6 hours of production data

Toy Story 2 (1998)¶

Engineer ran rm -rf * on the animation directory
Pixar nearly lost the entire film
Saved only because one employee had an informal copy on her home machine

Lesson: A backup that hasn't been tested is not a backup. It's a guess.

Backup Types¶

Type	How It Works	Use Case
Full backup	Copies everything	Weekly/monthly baseline
Incremental	Only changes since last backup	Daily — saves time and storage
Differential	Changes since last full backup	Faster restore than incremental
Snapshot	Point-in-time copy of a volume/DB	AWS EBS, RDS, FSx

AWS Backup Services¶

Service	What It Backs Up
S3 Versioning	Objects — keeps all versions, restores from any point
EBS Snapshots	Block storage volumes — incremental, stored in S3
RDS Automated Backups	Database — daily backup + transaction logs (point-in-time recovery)
RDS Snapshots	Manual point-in-time DB snapshot
AWS Backup	Centralised backup policy across EC2, EBS, RDS, DynamoDB, EFS
S3 Cross-Region Replication	Replicates objects to another region (DR)

Storage Classes (S3)¶

Frequency of access → cost tradeoff:

Standard           → frequent access, highest cost
Standard-IA        → infrequent access, lower cost, retrieval fee
One Zone-IA        → infrequent, single AZ (lower durability)
Glacier Instant    → archive, millisecond retrieval
Glacier Flexible   → archive, minutes to hours
Glacier Deep       → long-term archive, cheapest, 12h+ retrieval
Intelligent-Tiering → auto-moves objects based on access patterns

Mental Model: Storage Types in AWS¶

Need block storage (like a hard drive)?
  → EBS (attach to one EC2) or Instance Store (ephemeral, fast)

Need shared file system (multiple EC2s access same files)?
  → EFS (Linux) or FSx for Windows

Need object storage (files, backups, static assets)?
  → S3

Need data warehouse (analytics, SQL on PB of data)?
  → Redshift

Shell Profile Files (Side Note)¶

Mentioned during class — relevant when automating backup scripts:

File	When It Loads
`~/.bash_profile`	Login shells (SSH, console login)
`~/.bashrc`	Interactive non-login shells (new terminal tab)
`~/.zshrc`	Zsh equivalent of `.bashrc`

For scripts that run on boot or via cron, make sure environment variables are sourced from the right profile file.

AWS Context¶

S3 durability = 99.999999999% (11 nines) — data stored across multiple AZs
EBS — block storage, attached to one instance, persists independently of the instance
Instance Store — ephemeral, physically attached, lost on stop/terminate — don't store anything important
RDS point-in-time recovery — restore to any second within the backup retention window
S3 MFA Delete — require MFA to permanently delete versioned objects

30-Second Takeaway¶

Test your backups. A backup you've never restored is untested.
S3 = durable, versioned, cheap object storage
EBS = persistent block storage for EC2
Instance Store = fast but ephemeral — never for important data
AWS Backup = centralised policy management across services

Self-Quiz¶

What's the difference between incremental and differential backups?
Why did GitLab still lose data despite having 5 backup mechanisms?
Which S3 storage class would you use for logs accessed once a month?
What's the difference between EBS and Instance Store?
How does S3 versioning protect against accidental deletion?
What does 11 nines of durability actually mean?
Which AWS service centralises backup across EC2, RDS, and DynamoDB?