Phase Verbs / Action Phrases
| Phrase | Meaning |
|---|---|
| migrate (EBS across AZ) | snapshot → restore in new AZ |
| share (EFS across instances) | multiple instances mount the same EFS |
| lose (Instance Store data) | ephemeral data gone when instance stops/terminates |
| disable (Delete on Termination) | preserve root EBS volume after instance termination |
| leverage (storage tiers) | use EFS-IA or Archive to reduce costs |
Technical Vocabulary
| Term | Definition |
|---|---|
| EBS (Elastic Block Store) | Network-attached block storage; single instance per volume; AZ-scoped |
| EFS (Elastic File System) | Managed NFS; mounts to many instances across AZs; Linux only |
| Instance Store | Physical disk on the host server; fastest I/O; ephemeral (lost on stop/terminate) |
| POSIX | Standard Linux/Unix file API; EFS is POSIX-compliant |
| Mount target | The endpoint in each AZ through which instances connect to EFS |
Key Concepts
The Three EC2 Storage Types — Master Comparison
| EBS | EFS | Instance Store | |
|---|---|---|---|
| Type | Block storage | Network file system (NFS) | Physical disk |
| Attach to | 1 instance (io1/io2: up to 16) | Hundreds of instances | 1 instance (built-in) |
| AZ scope | Single AZ | Multi-AZ (or One Zone) | Same physical host |
| Persistence | Persistent | Persistent | Ephemeral (lost on stop) |
| OS support | Linux + Windows | Linux only | Linux + Windows |
| Capacity | Pre-provisioned | Auto-scaling | Fixed (hardware-defined) |
| Billing | Per GB provisioned | Per GB used | Included in instance cost |
| Performance | Up to 256,000 IOPS (io2) | 10 GB/s throughput | 3.3M IOPS (fastest) |
| Move across AZ | Via snapshot | Already multi-AZ | Cannot (hardware-bound) |
| Cost | Moderate | Higher (~3× gp2) | Lowest (built-in) |
EBS Key Points
- Locked to a specific AZ — cannot attach to instances in another AZ
- Migrating across AZ: snapshot → restore in target AZ
- gp2: IOPS linked to size; gp3/io1: IOPS independent of size
- EBS backups use I/O — don’t run backups during peak traffic
- Root EBS: deleted on termination by default (can disable)
EFS Key Points
- Attaches to hundreds of instances across different AZs simultaneously
- Mount targets in each AZ allow instances to connect
- Linux only (POSIX file system)
- Higher price than EBS, but pay per use + storage tiers save costs
- Great for: WordPress, web servers, shared content management
Instance Store Key Points
- Physically attached to the host machine — highest IOPS possible
- Data is ephemeral — lost when the instance stops or terminates
- Good for: temporary buffer, cache, scratch data, intermediate computation results
- Cannot be used for long-term storage
Choosing the Right Storage
| Scenario | Best Choice |
|---|---|
| Single EC2, persistent, need speed | EBS io1/io2 |
| Shared across multiple Linux instances | EFS |
| Maximum IOPS, temporary data | Instance Store |
| Database on a single instance | EBS gp3/io1 |
| WordPress shared media across web servers | EFS |
| Cheapest per GB, infrequent access | EFS-IA or sc1 EBS |
| Windows-based shared storage | NOT EFS — use other solutions |
Exam Tips
- EBS = 1 instance, 1 AZ (except Multi-Attach io1/io2)
- EFS = many instances, many AZs, Linux only
- Instance Store = fastest but ephemeral — data gone on stop/terminate
- EBS across AZ = snapshot required
- gp2: IOPS scales with disk size; gp3: IOPS and throughput are independent
- EFS is ~3× the cost of gp2 EBS but scales automatically and supports multi-AZ sharing
- EFS storage tiers (Standard → IA → Archive) can save up to 90%