Phase Verbs / Action Phrases
| Phrase | Meaning |
|---|---|
| mount (EFS to instances) | attach the network file system to EC2 instances |
| scale (automatically) | EFS grows/shrinks its capacity based on actual usage |
| move (files to storage tier) | lifecycle policy migrates infrequently accessed files to cheaper tier |
| provision (throughput) | set a fixed throughput independent of storage size |
| share (across AZs) | multiple instances in different AZs read/write the same EFS filesystem |
Technical Vocabulary
| Term | Definition |
|---|---|
| EFS (Elastic File System) | AWS managed NFS (Network File System) — a shared network drive for Linux EC2 instances |
| NFS (Network File System) | A protocol that allows files to be shared over a network (like a shared drive) |
| POSIX | A standard for Unix-like file APIs; EFS uses a standard POSIX-compatible API |
| Security Group | Controls access to the EFS mount points |
| Performance Mode | Setting that optimizes EFS for latency (General Purpose) or throughput (Max I/O) |
| Throughput Mode | How EFS delivers bandwidth: Bursting, Provisioned, or Elastic |
| Storage Tier | Where data is stored: Standard (frequent access), IA (infrequent), Archive (rare access) |
| EFS-IA | EFS Infrequent Access tier — cheaper storage, small fee to retrieve files |
| Lifecycle Policy | Rule that automatically moves files between storage tiers after N days of no access |
Key Concepts
What is EFS?
- Managed NFS — a network file system you can mount to many EC2 instances
- Unlike EBS (one instance per volume), EFS can be accessed by thousands of instances simultaneously
- Works across multiple AZs in the same region
- Linux only (POSIX-compatible) — NOT compatible with Windows
- No need to provision capacity; EFS scales automatically → pay per GB used
- ~3x more expensive than gp2 EBS, but eliminates capacity planning
EFS vs EBS Quick Comparison
| EFS | EBS | |
|---|---|---|
| Attach to | Many instances across AZs | One instance (same AZ) |
| OS | Linux only | Linux + Windows |
| Capacity | Auto-scales | Must provision |
| Billing | Per GB used | Per GB provisioned |
| Protocol | NFS | Block storage |
Performance Modes (set at creation)
| Mode | Latency | Throughput | Use Case |
|---|---|---|---|
| General Purpose (default) | Low | Moderate | Web servers, CMS, WordPress |
| Max I/O | Higher | Very High | Big data, media processing (highly parallel) |
Throughput Modes
| Mode | How It Works | Use Case |
|---|---|---|
| Bursting | Scales with storage size; can burst beyond baseline | Variable workloads |
| Provisioned | Set fixed throughput regardless of storage | Throughput needs exceed what storage size provides |
| Elastic | Auto-scales up/down based on actual I/O; up to 3 GB/s reads, 1 GB/s writes | Unpredictable workloads (recommended) |
Storage Tiers (Lifecycle Management)
| Tier | Access Frequency | Cost |
|---|---|---|
| Standard | Frequent | Higher storage cost |
| EFS-IA | Infrequent | Lower storage; retrieval fee |
| Archive | Rare (few times/year) | Lowest storage cost |
- Set up lifecycle policies to auto-move files after N days of no access
- Up to 90% cost savings by using the right tiers
Availability Modes
| Mode | AZ Coverage | Cost | Use Case |
|---|---|---|---|
| Standard (Multi-AZ) | Replicates across multiple AZs | Higher | Production workloads (HA) |
| One Zone | Single AZ; still has backups | Cheaper | Dev/test (EFS One Zone-IA available) |
Exam Tips
- EFS = NFS; mounts to many instances across AZs simultaneously
- Linux-only (POSIX) — not for Windows
- No capacity planning — pay per use (auto-scales)
- Performance modes: General Purpose (low latency) vs. Max I/O (high throughput, parallel)
- Throughput modes: Bursting / Provisioned / Elastic (for unpredictable)
- Storage tiers: Standard → EFS-IA → Archive; use lifecycle policies to move files
- Multi-AZ = production; One Zone = dev/test and cheaper