Learn
AWS
Storage
EFS

AWSGCP laterAzure later

EFSAmazon Elastic File System

StorageCore service · full depthFree while we build the libraryChecked against AWS docs · 15 Sept 2026

A shared NFS file system that grows and shrinks by itself, mounted at once by many instances, containers and functions across the zones of a Region.

In the examsSolutions Architect – AssociateHIGHSolutions Architect – ProfessionalMEDDeveloper – AssociateLOWSysOps Administrator – AssociateMED

Where EFS takes you6 steps, all open

I know the problem it solves
I can picture it
I know how it fails
I know when to choose it
I can use it in a design
I can answer an exam question

What & why

One file system, many clients

Amazon EFS is a file system you never size or manage. It grows and shrinks as files are added and removed, and many clients mount it at once over NFS: EC2 instances, containers on ECS, EKS and Fargate, and Lambda functions.

Regional or One Zone. A Regional file system, the kind AWS recommends, stores data across several Availability Zones and stays available when a zone is not. A One Zone file system keeps it in one zone, where losing the zone could lose the data. It works with Linux, not Windows.

Reach for it when

several instances must read and write the same files at once
an Auto Scaling group across zones needs shared content, such as uploads
storage should grow without anyone resizing it
containers or Lambda functions need a shared file system

It is the wrong tool when

one instance needs a low-latency disk, such as a database's: that is EBS
Windows servers need a file share: EFS does not support Windows
objects are served over HTTP, in any amount: that is S3

Mental model

The office's shared drive

The shared drive in an office: everyone on every floor sees the same folders, it never fills up, and you pay for what is on it. Each floor has its own socket to plug into. Files nobody has opened for a while move to a cheaper cabinet by themselves.

The shared driveone file system for many clientsEFS file system

A socket on each flooran address in each zone's subnetMount target

Plugging inmounting over NFSNFSv4.1

The cheaper cabinetfiles not read for 30 daysInfrequent Access

The basementfiles not read for 90 daysArchive

One building or severalwhere the data is keptRegional or One Zone

How it works

Create, add mount targets, mount

Create the file systemRegional or One Zone, with General Purpose performance and Elastic throughput: the defaults AWS recommends.
Add mount targetsOne in each Availability Zone, in a subnet, with a security group that decides which clients may connect.
Mount itClients mount it over NFS through the mount target in their own zone.
Let lifecycle tier itBy default, files not read for 30 days move to Infrequent Access, and after 90 days to Archive.

Protocol: NFSv4.1, NFSv4.0Linux clients
Latency: About 1 msreads; about 2.7 ms writes, Standard
Mount targets: One per zoneOne Zone: a single one
To Infrequent Access: 30 daysunread, by default
To Archive: 90 daysunread, by default
Replication: 15 min RPOfor most file systems

Architecture

Web servers that share their uploads

Every web server, in every zone, mounts the same file system, so a file uploaded through one is served by all of them.

A mount target in every zone the instances use. An instance that mounts through a mount target in another zone pays for data access across zones; access within the same zone is free.

A replica to fail over to. Replication keeps a copy in another Region, or the same one, and for most file systems changes reach it within 15 minutes.

Failure simulation

Lose a zone, Regional against One Zone

Regional file system, a mount target in each zone

AZ a

InstancesServed by Mount target a

AZ b

InstancesServed by Mount target b

Click a node to make it fail.

Everything is up. Take a node or a whole zone down.

One Zone file system in AZ a

AZ a

InstancesServed by Mount target a

AZ b

InstancesServed by Mount target a

Click a node to make it fail.

Everything is up. Take a node or a whole zone down.

Regional survives a zone. Instances in the zones that are still up keep reading and writing through their own mount targets. Only the instances in the lost zone go with it.

One Zone depends on one zone. Every instance goes through that zone's single mount target, the ones in other zones paying for access across zones. Losing the zone cuts all of them off, and the data itself could be lost, so back it up or replicate it.

Security

Who can connect, and what they can touch

Encrypt at rest when you create it. Encryption at rest is chosen when the file system is created, and then covers all its data and metadata.

Encrypt in transit when you mount it. Encryption of data in transit is turned on as the file system is mounted.

Two layers of access. IAM policies and network controls, such as the mount targets' security groups, decide which NFS clients can connect. POSIX permissions decide what they can do with each file.

Only root at first. On a new file system, only the root user on an instance can read, write and run files until you grant permissions.

Cost

How you are charged, not what it costs today

Storage used, not provisioned, per GB-month, lowest in Archive and highest in Standard.

Access charges for reads, writes and moving files between classes. With Elastic throughput you pay only for the throughput you use; Provisioned throughput is charged for what you set above the baseline your Standard storage includes.

Across zones. Access from another zone is charged as data transfer; from the same zone it is free.

Lifecycle saves the most on files that go unread: to Infrequent Access after 30 days and to Archive after 90, by default.

Prices change, so none are printed here. Check the EFS pricing page on aws.amazon.com.

Comparison

Where shared files belong

Storage	Kind	Used by	Where it lives	Choose it when
EFS	File system over NFS	Many Linux clients at once	A Region, or one zone	Shared files for instances, containers and functions
EBS	Block volume	One instance, or up to 16 with Multi-Attach	One Availability Zone	A disk for one instance, such as a database's
S3	Objects over HTTP	Any number of clients	A Region	Any amount of objects, websites and data lakes
Instance store	Block storage on the host	Its own instance only	The instance's host	Scratch data that can be lost

Throughput mode	How it scales	Choose it when
Elastic	Up and down with the workload; you pay for what you use	Spiky or unpredictable load: the default
Provisioned	A level you set, whatever the size	Known needs, with average load at 5 percent of peak or more
Bursting	With the amount in Standard storage, plus burst credits	Throughput should grow with what is stored

SAA traps

The phrase in the question, and what it means

instances in several Availability Zones must share the same filesAn EFS Regional file system

Each zone gets a mount target, and the data is stored across zones.

a shared file system for Windows serversNot EFS

EFS does not support Windows instances.

the file system must grow without anyone resizing itEFS

It grows and shrinks as files are added and removed.

cut the cost of files nobody has opened for a monthLifecycle management to Infrequent Access

Files unread in Standard for 30 days move there by default.

spiky, unpredictable throughputElastic throughput

It scales with the workload, and you pay only for what you use.

Max I/O performance mode, for more parallel clientsGeneral Purpose, almost always

Max I/O is the previous generation, with higher latency. AWS recommends General Purpose.

instances mount through a mount target in another zoneAdd a mount target in their own zone

Access across zones is charged; within a zone it is free.

a copy of the file system in another Region for disaster recoveryEFS Replication

For most file systems, changes reach the replica within 15 minutes.

turn on encryption at rest for an existing file systemCreate a new, encrypted file system

Encryption at rest is chosen when a file system is created.

Scenario

Exam style, marked when you answer

A company runs a content management system on Linux EC2 instances in an Auto Scaling group across three Availability Zones. Files uploaded through any instance must be visible to all of them at once. The storage must survive the loss of a zone and grow without anyone resizing it. What should a solutions architect use?

Practice

Fill the gap, the way the challenges work

Linux web servers in an Auto Scaling group across three zones must all read and write the same uploaded files at once, as an ordinary file system. The storage must grow by itself and stay available if a zone fails. What do the web servers mount?

ALB· Load balancerEC2 Auto Scaling· Web servers, three zonesPick a service below

Quick revision

The night before the exam

What: A shared NFS file system that grows and shrinks by itself.
Clients: EC2, ECS, EKS, Fargate and Lambda; Linux, never Windows.
Regional: Data across zones, a mount target in each: recommended.
One Zone: One zone and one mount target; the data can be lost with the zone.
Defaults: General Purpose performance, Elastic throughput.
Lifecycle: To Infrequent Access after 30 days, Archive after 90.
Replication: To another Region or the same, RPO 15 minutes.
Encryption: At rest chosen at creation; in transit when mounting.

This whole page is free right now.The AWS library is still being written, so every page of it is open to everyone while that lasts. It becomes a paid bundle later; what you read today costs you nothing.

Checked against

Every fact on this page was checked against AWS’s own documentation on 15 Sept 2026. If AWS has changed something since, its page is the one to trust.