EBS & EFS
AWS SAA SAA-C03 · Design High-Performing Architectures
Block storage: EBS volumes
Amazon Elastic Block Store (EBS) provides persistent block storage volumes that attach to EC2 instances in the same Availability Zone, like a virtual hard drive. An EBS volume exists independently of any instance: it survives when the instance stops, and it can be detached from one instance and attached to another in the same AZ. EBS volumes are automatically replicated within their AZ to protect against a single component failure, but they are AZ-scoped — a volume cannot be attached to an instance in a different AZ. That single fact drives many exam answers: if a design needs storage shared across AZs, EBS is the wrong service.
EBS volume types are the heart of this topic. General Purpose SSD (gp3 and gp2) is the default for most workloads: a balance of price and performance for boot volumes, development environments, and small-to-medium databases. gp3 lets you provision IOPS and throughput independently of volume size, with a baseline of 3,000 IOPS. Provisioned IOPS SSD (io2 and io1) is for I/O-intensive workloads that need consistently high performance, such as large databases — you pay for the IOPS you provision, up to very high levels. Throughput Optimized HDD (st1) is for frequently accessed, throughput-intensive workloads like big data and log processing, where sequential reads matter more than random IOPS. Cold HDD (sc1) is the cheapest option, for infrequently accessed data. The exam's decision rule: SSD for random I/O and latency-sensitive work, HDD for sequential throughput and cost-sensitive bulk data, and io2 when a workload needs sustained high IOPS that gp3 cannot deliver.
Snapshots: backup and portability
An EBS snapshot is a point-in-time backup of a volume, stored in S3. Snapshots are incremental: only the blocks that changed since the previous snapshot are stored, which keeps them cheap and fast. You can create a new volume from a snapshot in the same Region, copy snapshots to other Regions (which is how you move EBS data across Regions), and share snapshots with other accounts. Snapshots are also the mechanism behind disaster recovery for EBS-backed systems: a scheduled snapshot policy (through Amazon Data Lifecycle Manager) gives you regular recovery points, and you restore by creating volumes from the snapshots. Snapshots of encrypted volumes stay encrypted, and you can enable encryption by default so that every new volume and snapshot is encrypted automatically with a KMS key.
A related concept is the instance store: physically attached, ephemeral storage on the EC2 host. Instance store volumes offer very high IOPS and low latency, but their data is lost if the instance stops, terminates, or the host fails. They are right for temporary data that can be regenerated — caches, buffers, scratch space — and wrong for anything durable. The exam often contrasts instance store with EBS: if the data must survive a stop or a host failure, use EBS.
Shared file storage: EFS
Amazon Elastic File System (EFS) is a managed Network File System (NFS) that many EC2 instances can mount at the same time, across multiple Availability Zones in a Region. It grows and shrinks automatically as files are added and removed, so there is no capacity to provision. EFS is the exam's answer whenever multiple instances in different AZs need to read and write the same files — a shared content directory for a fleet of web servers, for example. EFS offers storage classes (Standard and Infrequent Access, with lifecycle management to move files automatically) and two performance modes: General Purpose for latency-sensitive workloads, and Max I/O for highly parallel workloads that can tolerate slightly higher latency. Throughput modes include Bursting (the default, which scales with the amount of data stored) and Provisioned (for workloads needing a fixed throughput regardless of size).
AWS also offers Amazon FSx for workload-specific file systems: FSx for Windows File Server (native Windows SMB file shares, integrated with Active Directory), FSx for Lustre (high-performance computing), FSx for NetApp ONTAP, and FSx for OpenZFS. The exam mainly tests FSx for Windows File Server, which is the answer when a Windows application needs a shared SMB file system. The key comparison: EFS is for Linux NFS workloads, FSx for Windows is for Windows SMB workloads, and EBS is for a single instance's block storage.
Choosing between EBS, EFS, instance store, and S3
| Need | Best choice |
|---|---|
| Block storage for one instance, must survive stops | EBS |
| Shared files across many instances and AZs (Linux) | EFS |
| Shared files for Windows applications (SMB) | FSx for Windows File Server |
| Temporary high-speed scratch data, loss is acceptable | Instance store |
| Object storage, static files, backups, archives | S3 |
The exam loves this table in question form. A question will describe a workload — number of instances, operating system, durability need, access pattern — and exactly one storage service will fit all the constraints. Read the scenario for the constraints (shared? multi-AZ? Windows? durable?) before looking at the options.
Performance tuning for block and file storage
Getting the best performance from EBS starts with matching the volume type to the workload, but there are further levers. gp3 volumes let you provision IOPS and throughput independently of volume size — you can raise IOPS up to 16,000 and throughput up to 1,000 MiB/s without buying a larger volume, which makes gp3 the flexible default for most workloads. io2 Block Express is the highest-performance EBS option, delivering up to 256,000 IOPS per volume for the most demanding database workloads on supported instance types. EBS Multi-Attach lets multiple instances in the same Availability Zone attach to a single io1 or io2 volume at the same time, which is useful for clustered applications that manage shared block access themselves — note that Multi-Attach does not make EBS a shared file system, and the application must handle concurrent writes. EBS Fast Snapshot Restore eliminates the first-access latency when creating a volume from a snapshot, so restored volumes perform at full speed immediately instead of warming up lazily.
For EFS, performance tuning means choosing the right modes. General Purpose performance mode suits latency-sensitive workloads such as web serving and content management, while Max I/O suits highly parallel workloads that can tolerate slightly higher per-operation latency. Bursting throughput mode scales the available throughput with the amount of data stored, which works well when the file system is large; Provisioned throughput mode guarantees a fixed throughput regardless of size, which is the answer when a small file system needs high throughput. EFS lifecycle management automatically moves files that have not been accessed for a configured period to the Infrequent Access storage class, cutting costs for file systems with a long tail of cold data. For high-performance computing workloads that need a parallel file system, Amazon FSx for Lustre provides sub-millisecond latencies and throughput that scales with the size of the file system, and it integrates directly with S3 so data can be processed in place.
Exam traps to watch for
- Using EBS for storage that must be shared across Availability Zones. EBS volumes live in one AZ; shared multi-AZ file access is EFS.
- Storing durable data on instance store volumes. Instance store data is lost when the instance stops, terminates, or the host fails.
- Choosing gp2/gp3 for a workload that needs sustained very high IOPS. Provisioned IOPS (io2) exists for exactly that case.
- Forgetting snapshots are incremental and Regional: copying a snapshot is how you move EBS data to another Region, and restoring means creating a new volume from the snapshot.
Practice questions
Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.
Q1.A company runs a fleet of web servers in two Availability Zones. All servers need read and write access to the same shared directory of uploaded images. Which storage solution meets this requirement?
Explanation: EFS is a shared NFS file system that many instances can mount simultaneously across Availability Zones, which is exactly what a shared image directory needs. Option D gives each server its own separate volume, so uploads to one server would be invisible to the others. Option C is ephemeral per-instance storage that is lost on stop and is not shared. Option B is not possible: an EBS volume can only attach to instances in its own Availability Zone.
Exam objective: SAA-C03 Task Statement 3.1: Determine high-performing and/or scalable storage solutions
Q2.A database workload on an EC2 instance needs sustained random I/O performance of 20,000 IOPS with low latency. The current gp3 volume cannot deliver consistent performance at that level. Which EBS volume type should the solutions architect choose?
Explanation: Provisioned IOPS SSD (io2) is designed for I/O-intensive workloads that need sustained, predictable high IOPS — you provision the IOPS you need and the volume delivers them consistently. Option C and B are HDD types optimized for sequential throughput, not random IOPS, and would be far too slow. Option A is the older general-purpose type; gp2's performance is tied to volume size and cannot guarantee 20,000 sustained IOPS.
Exam objective: SAA-C03 Task Statement 3.1: Determine high-performing and/or scalable storage solutions
Q3.A company needs a disaster recovery copy of its EBS volumes in a second AWS Region. Which approach achieves this?
Explanation: Snapshots are the portable form of EBS data: create a snapshot, copy it to the second Region, and create new volumes from it there when needed. Option B is impossible because EBS volumes are AZ-scoped and cannot attach across Regions. Option C misstates Multi-Attach, which lets multiple instances in the SAME Availability Zone share one io1/io2 volume — it does not span Regions. Option D destroys durability: instance store data is ephemeral and cannot be copied as a backup.
Exam objective: SAA-C03 Task Statement 2.2: Design highly available and/or fault-tolerant architectures
Q4.A video transcoding application uses EC2 instances with fast local scratch disks for temporary working files. The files are re-created from source data if lost, so durability of the scratch disks does not matter. Which storage option provides the highest I/O performance at the lowest cost for this use case?
Explanation: Instance store volumes are physically attached to the host, giving very high IOPS and low latency, and they are included in the price of the instance. Since the scratch data is temporary and re-creatable, the loss of data on stop or host failure is acceptable. Option B works but is expensive persistent storage for data that does not need to persist. Option C is shared network file storage, slower than local disks and charged separately. Option D is object storage accessed over the network, unsuitable as a local scratch disk.
Exam objective: SAA-C03 Task Statement 3.1: Determine high-performing and/or scalable storage solutions
Q5.A company is migrating a Windows-based file server to AWS. The application requires a native Windows SMB file share integrated with the company's Active Directory. Which storage service should the solutions architect recommend?
Explanation: FSx for Windows File Server provides fully managed native Windows file shares over SMB with Active Directory integration — the direct replacement for an on-premises Windows file server. Option C is NFS file storage for Linux workloads and does not provide native SMB/Active Directory integration. Option B is object storage, not a file system, and applications expecting SMB shares cannot use it directly. Option D is block storage for a single instance, not a shared Windows file service.
Exam objective: SAA-C03 Task Statement 3.1: Determine high-performing and/or scalable storage solutions
Q6.A solutions architect wants to encrypt all new EBS volumes and snapshots in an AWS account automatically, without relying on developers to remember the setting. What should the architect do?
Explanation: EBS encryption by default is a Region-level setting: once enabled, every new volume and every snapshot copy is encrypted automatically, with no action needed from developers. Option C can block unencrypted volumes but does not encrypt anything — it only causes failures. Option A encrypts S3 objects, not EBS volumes. Option D detects the problem after the fact instead of preventing it.
Exam objective: SAA-C03 Task Statement 1.3: Determine appropriate data security controls
Q7.A company runs a big data analytics workload that performs large sequential reads over hundreds of gigabytes of data. The workload is throughput-sensitive but not latency-sensitive, and cost matters. Which EBS volume type is MOST appropriate?
Explanation: Throughput Optimized HDD (st1) is built for frequently accessed, throughput-intensive workloads with large sequential I/O — big data, log processing, data warehousing — at a lower price per gigabyte than SSD. Option D is optimized for random IOPS, which this workload does not need, and costs much more. Option B would work but costs more per GB than st1 for a workload that does not need SSD latency. Option C is for infrequently accessed data; this workload reads its data frequently.
Exam objective: SAA-C03 Task Statement 4.1: Design cost-optimized storage solutions
Q8.An EC2 instance is terminated accidentally. The instance used an EBS-backed root volume with the default delete-on-termination setting, and no snapshots exist. What happens to the data on the root volume?
Explanation: By default, the root EBS volume of an instance is deleted when the instance is terminated, so without snapshots the data is gone. Option D describes what happens when an instance is STOPPED (or when delete-on-termination is disabled), not terminated with defaults. Option A invents an automatic backup that does not exist. Option B is wrong: the default delete-on-termination behavior for root volumes deletes them on termination (additional non-root volumes default to being preserved).
Exam objective: SAA-C03 Task Statement 2.2: Design highly available and/or fault-tolerant architectures