What the exam asks
- Pick the EBS volume type from an IOPS, throughput, latency or durability need and the qualifier, usually MOST cost-effective.
- Recognise when instance store is right: temporary, rebuildable data that needs the highest random I/O.
- Choose between EFS and the four FSx file systems by protocol, operating system, directory integration and workload.
- Tell block, file and object storage apart, and choose the one that scales to the stated future need.
- Find the hidden bottleneck: gp2 burst credits, the instance’s EBS bandwidth, the EFS throughput mode, or AZ scope.
Core ideas
Block, file or object?
| Storage | Access | Scope | Shared? |
|---|---|---|---|
| EBS | Block device | One AZ | No (Multi-Attach io1/io2 only, same AZ) |
| Instance store | Local block device on the host | That host | No |
| EFS | NFS v4.x, POSIX, Linux | Regional (multi-AZ) or One Zone | Thousands of clients |
| FSx | SMB, NFS, Lustre, iSCSI | Single-AZ or Multi-AZ | Yes |
| S3 | HTTPS object API | Regional | Yes, unlimited scale |
“Shared”, “concurrently” or “multiple AZs” rules out EBS and instance store. “Grows to petabytes” plus HTTP access means S3.
EBS volume types
| Type | Built for | Key behaviour | Exam trigger |
|---|---|---|---|
| gp3 (SSD) | Most workloads | 3,000 IOPS and 125 MiB/s baseline at any size. Extra IOPS and throughput are provisioned . Cheapest SSD |