Block and File Storage: EBS, EFS and Amazon FSx Deep Dive#

S3 (Simple Storage Service) handles objects, but applications running on EC2 (Elastic Compute Cloud) or containers often need block storage (like a local hard drive) or shared file storage (NFS/SMB). Choosing the right storage type, volume tier, or FSx flavor is one of the most heavily tested topics on the SAA-C03 exam.

Block vs File vs Object at a glance#

flowchart TB
    s{Storage Requirement?}
    s -->|Single EC2 / low-latency DB disk| ebs[EBS / Instance Store]
    s -->|Shared Linux POSIX across AZs| efs[Amazon EFS]
    s -->|Windows SMB / AD share| fsxw[FSx for Windows]
    s -->|HPC / ML + S3 integration| fsxl[FSx for Lustre]
    s -->|Enterprise NetApp / multi-protocol| fsxn[FSx for NetApp ONTAP]
    s -->|HTTP REST / static / archive| s3[Amazon S3]

Amazon EBS (Elastic Block Store)#

EBS volumes are network-attached block disks locked to a single Availability Zone (AZ). To move a volume to another AZ, you snapshot it and restore in the target AZ.

EBS Volume Types#

Type	Category	Max IOPS / Throughput	Exam Use Case
gp3 / gp2	General Purpose SSD	16,000 IOPS (gp3 scales IOPS & throughput independently of size!)	Boot volumes, dev/test, most workloads
io1 / io2 / io2 Block Express	Provisioned IOPS SSD	64,000 – 256,000 IOPS, 99.999% durability (io2)	Mission-critical databases needing >16,000 IOPS or sub-millisecond latency
st1	Throughput Optimized HDD	500 IOPS / 500 MB/s	Big data, data warehouses, sequential log processing (CANNOT be a boot volume)
sc1	Cold HDD	250 IOPS / 250 MB/s	Cheapest block storage for infrequent sequential access (CANNOT be a boot volume)

EBS Exam Traps & Features#

EBS Multi-Attach: Available only on io1 and io2 within a single AZ (up to 16 EC2 instances). You must use a cluster-aware file system (like GFS2); standard EXT4/XFS will corrupt data! If the question wants a simple shared file system across AZs, pick EFS, not Multi-Attach.
Instance Store vs EBS: Instance Store is physically attached NVMe/SSD on the host → highest possible IOPS (hundreds of thousands), but ephemeral (data is lost if the instance stops or the host fails; survives reboots). Use for temporary caches, buffers, or scratch data.
Snapshots:
- Incremental backups stored in S3.
- EBS Snapshot Archive: up to 75% cheaper for snapshots kept 90+ days; takes 24–72 hours to restore.
- Fast Snapshot Restore (FSR): eliminates the initial I/O latency penalty (“pre-warming”) on restored volumes (costs extra per hour).
- Recycle Bin for EBS Snapshots & AMIs (Amazon Machine Image): protects against accidental deletion with retention rules.
RAID on EBS: RAID 0 stripes across 2+ volumes to multiply IOPS/throughput (no fault tolerance); RAID 1 mirrors for redundancy (rarely needed since EBS replicates within the AZ).

Amazon EFS (Elastic File System)#

Managed NFSv4 (Linux/POSIX) shared file system that scales automatically and can be mounted by thousands of EC2 instances, Lambda functions, and ECS (Elastic Container Service)/EKS (Elastic Kubernetes Service) containers across multiple AZs.

Knob	Options	When to pick
Storage Classes	Standard (Multi-AZ) vs One Zone (1 AZ, ~47% cheaper) + IA (Infrequent Access) via Lifecycle	Production shared files → Standard + Lifecycle to IA; Dev/test or reproducible → EFS One Zone
Performance Mode	General Purpose (default, lowest latency) vs Max I/O (higher aggregate throughput, slightly higher latency)	Web serving / CMS → General Purpose; Big data / media processing across 100s of nodes → Max I/O
Throughput Mode	Elastic (auto-scales for unpredictable traffic), Bursting (scales with storage size), Provisioned (set fixed MB/s regardless of size)	Small dataset needing high throughput → Provisioned; Spiky/unknown → Elastic

Sınav Tuzağı (Exam Trap): EFS sadece Linux (POSIX/NFS) destekler; Windows EC2 instance’ları için yerel SMB paylaşımı sorulursa cevap EFS değil, FSx for Windows File Server‘dır!

Amazon FSx - the 4 specialized file systems#

Whenever EFS doesn’t fit because of protocol (Windows/SMB), extreme HPC speed, or enterprise storage features, pick the matching FSx service:

Service	Protocol / OS	Key Exam Keywords & Features
FSx for Windows File Server	SMB & NTFS (Windows)	Active Directory integration, ACLs, DFS (Distributed File System) Namespaces, Multi-AZ HA, SQL Server failover clusters
FSx for Lustre	POSIX (Linux)	HPC (High Performance Computing), Machine Learning, sub-ms latency, hundreds of GB/s. Seamless S3 integration (presents an S3 bucket as a fast file system and writes results back). Modes: Scratch (temporary, bursty, no replication, cheaper) vs Persistent (replicated within the AZ)
FSx for NetApp ONTAP	Multi-protocol (NFS, SMB, iSCSI)	Migrate existing NetApp / ONTAP workloads; works with Linux, Windows & macOS simultaneously; built-in SnapMirror, instant clones, compression & deduplication
FSx for OpenZFS	NFS (v3/v4)	Migrate ZFS file servers; point-in-time snapshots, copy-on-write clones, sub-millisecond latency

Exam check#

Q1: A machine learning team stores 80 TB of training datasets in Amazon S3. Their GPU cluster on EC2 needs to process this data as a POSIX file system with hundreds of GB/s throughput for a 6-hour training run, at the lowest cost. If a storage server fails during the run, the job can simply be restarted. Which storage solution should be used?

A) Amazon EFS with Max I/O performance mode
B) Amazon FSx for Lustre with a Scratch file system linked to the S3 bucket
C) Amazon FSx for Lustre with a Persistent file system
D) EBS gp3 volumes with Multi-Attach

Cevabı göster

Answer: B. HPC/ML + native S3 integration + hundreds of GB/s = FSx for Lustre. Because the job is short-lived and can tolerate failure (“lowest cost”), Scratch beats Persistent.

Q2: A legacy relational database running on EC2 requires a single block storage volume capable of sustaining 50,000 IOPS consistently. Which storage option meets this requirement?

A) EBS gp3
B) EBS io2 (Provisioned IOPS SSD)
C) EBS st1 (Throughput Optimized HDD)
D) Amazon EFS Standard

Cevabı göster

Answer: B. gp3 tops out at 16,000 IOPS per volume. Sustained block IOPS above 16,000 (up to 64,000–256,000) requires io1/io2.

Q3: A company runs a Windows-based .NET application across multiple AZs that requires shared SMB storage integrated with their on-premises Microsoft Active Directory. Which service should the architect choose?

A) Amazon EFS Multi-AZ
B) Amazon S3 with File Gateway
C) Amazon FSx for Windows File Server (Multi-AZ)
D) EBS Multi-Attach with io2

Cevabı göster

Answer: C. Shared native Windows SMB + Active Directory + Multi-AZ = FSx for Windows File Server.

Ozet#

EBS = tek AZ’ye bağlı blok disk; gp3 (16K IOPS’a kadar, bağımsız ölçek), io2 (>16K IOPS, kritik DB, Multi-Attach), st1/sc1 (HDD, boot olamaz)
Instance Store = fiziksel sunucu üzerindeki NVMe disk; en yüksek IOPS ama geçici (instance durursa veri silinir)
EBS Multi-Attach = sadece io1/io2, tek AZ, cluster-aware dosya sistemi şart
EFS = çoklu AZ paylaşımlı Linux (NFS/POSIX) dosya sistemi; maliyet için EFS One Zone ve IA Lifecycle
FSx for Windows = SMB, NTFS, Active Directory, Multi-AZ
FSx for Lustre = HPC/ML, S3 entegrasyonu; Scratch (ucuz/geçici) vs Persistent (kalıcı)
FSx for NetApp ONTAP = çoklu protokol (NFS+SMB+iSCSI), deduplication, SnapMirror
FSx for OpenZFS = ZFS göçü, anlık klon ve snapshot