While preparing for the AWS SAA-C03, many candidates get confused by storage persistence terminology. In the real world, this is fundamentally a decision about durability guarantees vs. access latency. Let’s drill into a simulated scenario.
The Scenario #
GreenLeaf Marketplace, an e-commerce platform selling organic products, currently stores its product catalog (images, descriptions, pricing metadata) directly on EC2 instance store volumes attached to their web application servers. During a recent instance maintenance event, the entire catalog was lost, causing a 4-hour outage and $80,000 in lost revenue.
The engineering director has mandated:
- Zero data loss during instance replacements or failures
- High availability across multiple Availability Zones
- Minimal application refactoring (the app currently uses POSIX file system calls)
Key Requirements #
Migrate the product catalog to a persistent, highly available storage solution that survives EC2 instance terminations while maintaining compatibility with existing file-based application code.
The Options #
- A) Migrate the product catalog to Amazon ElastiCache for Redis.
- B) Deploy larger EC2 instances with larger instance store volumes.
- C) Migrate the product catalog from instance store to Amazon S3 Glacier Deep Archive.
- D) Migrate the product catalog to Amazon Elastic File System (Amazon EFS).
Correct Answer #
Option D.
The Architect’s Analysis #
Correct Answer #
Option D: Amazon EFS
Step-by-Step Winning Logic #
Amazon EFS provides:
- Durability: 99.999999999% (11 nines) with automatic replication across multiple AZs
- High Availability: Regional service—survives AZ failures without manual intervention
- POSIX Compatibility: Drop-in replacement for instance store—no application refactoring required
- Elastic Scalability: Automatically grows/shrinks based on data volume
- Shared Access: Multiple EC2 instances can mount the same file system simultaneously (critical for auto-scaling groups)
The Technical Win: EFS directly addresses all three requirements—persistence, HA, and file system compatibility—with a single managed service.
The Traps (Distractor Analysis) #
Why not Option A (ElastiCache for Redis)? #
- Wrong Abstraction Layer: Redis is an in-memory key-value store, not a file system
- Application Rewrite Required: Would need to refactor all file I/O calls to Redis API commands
- Cost Inefficiency: Paying for memory pricing (~$0.068/GB-hour for cache.r6g.large) for cold product data
- Persistence Complexity: Redis persistence (RDB/AOF) is designed for cache recovery, not source-of-truth durability
Exam Trap: ElastiCache is the correct answer when the question asks about low-latency session state or database query caching, not file storage.
Why not Option B (Larger Instance Store)? #
- Fundamentally Ephemeral: Instance store data is lost on stop/terminate/hardware failure regardless of volume size
- No Multi-AZ Resilience: Instance store is physically attached to the host—single point of failure
- Violates Core Requirement: Directly contradicts the “persistent location” mandate
Exam Trap: AWS will never recommend instance store for data requiring durability—this is a “distractor by amplification” (suggesting more of the wrong thing).
Why not Option C (S3 Glacier Deep Archive)? #
- Retrieval Latency: 12-48 hours to restore data—catastrophic for a live product catalog
- Wrong Access Pattern: Glacier Deep Archive is for compliance archives accessed <1/year, not active web content
- Cost Structure Mismatch: Cheapest storage ($0.00099/GB-month) but retrieval costs $0.02/GB + wait time
Exam Trap: Any Glacier option is wrong unless the question explicitly mentions “regulatory retention,” “long-term archive,” or “infrequent access.”
The Architect Blueprint #
Diagram Note: All EC2 instances in the Auto Scaling Group mount the same EFS file system via NFS across multiple AZ mount targets, ensuring the product catalog survives individual instance or AZ failures while remaining accessible to all active servers.
The Decision Matrix #
| Option | Est. Complexity | Est. Monthly Cost (10 TB Catalog) | Pros | Cons |
|---|---|---|---|---|
| A) ElastiCache Redis | High (API rewrite) | ~$4,900 (cache.r6g.4xlarge cluster) | Sub-ms latency; advanced data structures | Not a file system; volatile memory; expensive for cold data |
| B) Larger Instance Store | Low (no change) | $0 (included in instance) | Highest I/O performance; no storage cost | Ephemeral; no HA; data loss on stop/terminate |
| C) S3 Glacier Deep Archive | Medium (S3 API) | ~$10 (storage only) | Lowest storage cost; 99.999999999% durability | 12-48h retrieval; unusable for live catalog |
| D) Amazon EFS ✅ | Low (POSIX mount) | ~$3,000 (Standard class) or ~$250 (IA class with Lifecycle) | POSIX compatible; multi-AZ HA; managed; elastic | Higher latency than instance store; costlier than S3 |
FinOps Insight: For frequently accessed data (>50% of catalog), EFS Standard at $0.30/GB-month is justified. With EFS Intelligent-Tiering, infrequently accessed products automatically move to IA class ($0.025/GB-month), reducing costs by ~90% for cold inventory.
Real-World Practitioner Insight #
Exam Rule #
For the SAA-C03 exam:
- If the question mentions “persistent,” “durable,” or “survives instance termination” + file system operations, choose EBS (single instance) or EFS (multi-instance/HA).
- Instance store is ONLY correct for ephemeral workloads: swap space, temp files, caches, buffers.
Real World #
In production at scale, we’d likely implement a hybrid architecture:
- Source of Truth: Store the master catalog in Amazon S3 (11 nines durability, $0.023/GB-month, versioning enabled)
- Application Layer: Mount EFS as a working cache with CloudWatch + Lambda to sync from S3 on file changes
- Edge Acceleration: Use Amazon CloudFront to cache product images/metadata at 450+ global edge locations
- Cost Optimization: Implement S3 Intelligent-Tiering to automatically move stale product data to Infrequent Access ($0.0125/GB-month)
Why the Exam Doesn’t Test This: SAA-C03 focuses on service selection fundamentals, not multi-tier optimization (that’s SAP-C02 territory). The exam wants you to recognize that EFS solves the immediate POSIX + HA requirement without over-engineering.
The $80K Lesson: The original outage cost equals 27 months of EFS Standard storage for 10 TB. Durability isn’t an operational expense—it’s revenue insurance.