Storage & S3AWS-03 · theory

Source · AWS SAA-C03 Exam Guide + AWS Storage documentation

Why this matters

SAA-C03 Exam Guide, Domains 3 & 4

Storage choices drive both performance and cost, the second and fourth exam domains. Choosing S3 Standard when Glacier would do can multiply a bill; choosing EBS when EFS was needed can break a multi-instance design.

The exam constantly asks you to match a durability, access pattern, and cost requirement to exactly one storage service and class.

The concept

AWS S3 User Guide

Amazon S3 is object storage with eleven nines (99.999999999%) of durability by design. Its storage classes trade retrieval speed and cost: S3 Standard (frequent access), S3 Standard-IA and One Zone-IA (infrequent, lower storage cost + retrieval fee), S3 Intelligent-Tiering (auto-moves objects by access pattern), and the Glacier family — Glacier Instant Retrieval, Glacier Flexible Retrieval, and Glacier Deep Archive — for archival at the lowest cost with longer retrieval times.

Lifecycle rules automatically transition objects between classes and expire them, so you don't overpay for cold data.

Block and file storage differ fundamentally. EBS is a network block volume attached to one EC2 instance (in one AZ). EFS is an elastic NFS file system that many instances across AZs can mount concurrently. S3 is not a filesystem at all — it's accessed over HTTP APIs.

Worked scenario

AWS EFS and S3 Glacier documentation

A fleet of EC2 instances across three AZs must share a common set of files with POSIX semantics and read/write concurrency. Which storage?

The answer is EFS, because it can be mounted by many instances across AZs simultaneously. EBS is wrong — a standard EBS volume attaches to a single instance in a single AZ. S3 is wrong — it lacks POSIX file semantics.

Now flip it: archive compliance logs accessed maybe once a year, retrieval within 12 hours acceptable, lowest cost. That's S3 Glacier Deep Archive, ideally reached via a lifecycle rule that transitions logs there after 90 days.

How it connects

AWS Well-Architected Framework

Storage overlaps heavily with cost optimization — lifecycle policies and Intelligent-Tiering are core cost-saving tools. It also touches security: S3 supports encryption at rest (SSE-S3, SSE-KMS) and access control via bucket policies and Block Public Access.

Durability versus availability is a recurring distinction: S3 Standard offers 11 nines durability but 99.99% availability; One Zone-IA keeps the same durability within one AZ but loses data if that AZ is destroyed. The exam loves this trade-off.

Common traps
  • S3 durability (11 nines) is not the same as availability (99.99% for Standard).
  • One Zone-IA stores in a single AZ — cheaper, but data is lost if that AZ is destroyed.
  • A standard EBS volume attaches to ONE instance in ONE AZ; use EFS for shared multi-AZ access.
Key takeaways
  • Match access pattern to S3 class; automate transitions with lifecycle rules.
  • EBS = single-instance block; EFS = multi-instance file; S3 = object over HTTP.
  • Glacier Deep Archive = lowest cost, longest retrieval, for rarely accessed archives.