I’m excited to announce Amazon S3 Recordsdata, a brand new file system that seamlessly connects any AWS compute useful resource with Amazon Easy Storage Service (Amazon S3).
Greater than a decade in the past, as an AWS coach, I spent numerous hours explaining the elemental variations between object storage and file methods. My favourite analogy was evaluating S3 objects to books in a library (you may’t edit a web page, you want to exchange the entire guide) versus recordsdata in your pc that you could modify web page by web page. I drew diagrams, created metaphors, and helped clients perceive why they wanted completely different storage varieties for various workloads. Properly, right this moment that distinction turns into a bit extra versatile.
With S3 Recordsdata, Amazon S3 is the primary and solely cloud object retailer that provides fully-featured, high-performance file system entry to your knowledge. It makes your buckets accessible as file methods. This implies modifications to knowledge on the file system are mechanically mirrored within the S3 bucket and you’ve got fine-grained management over synchronization. S3 Recordsdata will be connected to a number of compute sources enabling knowledge sharing throughout clusters with out duplication.
Till now, you had to decide on between Amazon S3 value, sturdiness, and the providers that may natively eat knowledge from it or a file system’s interactive capabilities. S3 Recordsdata eliminates that tradeoff. S3 turns into the central hub for all of your group’s knowledge. It’s accessible instantly from any AWS compute occasion, container, or operate, whether or not you’re operating manufacturing functions, coaching ML fashions, or constructing agentic AI methods.
You may entry any common goal bucket as a local file system in your Amazon Elastic Compute Cloud (Amazon EC2) situations, containers operating on Amazon Elastic Container Service (Amazon ECS) or Amazon Elastic Kubernetes Service (Amazon EKS), or AWS Lambda features. The file system presents S3 objects as recordsdata and directories, supporting all Community File System (NFS) v4.1+ operations like creating, studying, updating, and deleting recordsdata.
As you’re employed with particular recordsdata and directories via the file system, related file metadata and contents are positioned onto the file system’s high-performance storage. By default, recordsdata that profit from low-latency entry are saved and served from the excessive efficiency storage. For recordsdata not saved on excessive efficiency storage corresponding to these needing giant sequential reads, S3 Recordsdata mechanically serves these recordsdata instantly from Amazon S3 to maximise throughput. For byte-range reads, solely the requested bytes are transferred, minimizing knowledge motion and prices.
The system additionally helps clever pre-fetching to anticipate your knowledge entry wants. You even have fine-grained management over what will get saved on the file system’s excessive efficiency storage. You may resolve whether or not to load full file knowledge or metadata solely, which implies you may optimize on your particular entry patterns.
Underneath the hood, S3 Recordsdata makes use of Amazon Elastic File System (Amazon EFS) and delivers ~1ms latencies for lively knowledge. The file system helps concurrent entry from a number of compute sources with NFS close-to-open consistency, making it very best for interactive, shared workloads that mutate knowledge, from agentic AI brokers collaborating via file-based instruments to ML coaching pipelines processing datasets.
Let me present you easy methods to get began.
Creating my first Amazon S3 file system, mounting, and utilizing it from an EC2 occasion is easy.
I’ve an EC2 occasion and a common goal bucket. On this demo, I configure an S3 file system and entry the bucket from an EC2 occasion, utilizing common file system instructions.
For this demo, I take advantage of the AWS Administration Console. It’s also possible to use the AWS Command Line Interface (AWS CLI) or infrastructure as code (IaC).
Right here is the structure diagram for this demo.
Step 1: Create an S3 file system.
On the Amazon S3 part of the console, I select File methods after which Create file system.
I enter the identify of the bucket I need to expose as a file system and select Create file system.
Step 2: Uncover the mount goal.
A mount goal is a community endpoint that can dwell in my digital personal cloud (VPC). It permits my EC2 occasion to entry the S3 file system.
The console creates the mount targets mechanically. I take notes of the Mount goal IDs on the Mount targets tab.
When utilizing the CLI, two separate instructions are essential to create the file system and its mount targets. First, I create the S3 file system with create-file-system. Then, I create the mount goal with create-mount goal.
Step 3: Mount the file system on my EC2 occasion.
After it’s linked to an EC2 occasion, I sort:
sudo mkdir /residence/ec2-user/s3files sudo mount -t s3files fs-0aa860d05df9afdfe:/ /residence/ec2-user/s3files
I can now work with my S3 knowledge instantly via the mounted file system in ~/s3files, utilizing commonplace file operations.
After I make updates to my recordsdata within the file system, S3 mechanically manages and exports all updates as a brand new object or a brand new model on an current object again in my S3 bucket inside minutes.
Modifications made to things on the S3 bucket are seen within the file system inside just a few seconds however can generally take a minute or longer.
# Create a file on the EC2 file system
echo "Hello S3 Files" > s3files/hi there.txt
# and confirm it is right here
ls -al s3files/hi there.txt
-rw-r--r--. 1 ec2-user ec2-user 15 Oct 22 13:03 s3files/hi there.txt
# See? the file can also be on S3
aws s3 ls s3://s3files-aws-news-blog/hi there.txt
2025-10-22 13:04:04 15 hi there.txt
# And the content material is equivalent!
aws s3 cp s3://s3files-aws-news-blog/hi there.txt . && cat hi there.txt
Whats up S3 RecordsdataIssues to know
Let me share some vital technical particulars that I believe you’ll discover helpful.
One other query I steadily hear in buyer conversations is about selecting the best file service on your workloads. Sure, I do know what you’re considering: AWS and its seemingly overlapping providers, retaining cloud architects entertained throughout their structure assessment conferences. Let me assist demystify this one.
S3 Recordsdata works finest once you want interactive, shared entry to knowledge that lives in Amazon S3 via a excessive efficiency file system interface. It’s very best for workloads the place a number of compute sources—whether or not manufacturing functions, agentic AI brokers utilizing Python libraries and CLI instruments, or machine studying (ML) coaching pipelines—must learn, write, and mutate knowledge collaboratively. You get shared entry throughout compute clusters with out knowledge duplication, sub-millisecond latency, and automated synchronization together with your S3 bucket.
For workloads migrating from on-premises NAS environments, Amazon FSx supplies the acquainted options and compatibility you want. Amazon FSx can also be very best for high-performance computing (HPC) and GPU cluster storage with Amazon FSx for Lustre. It’s notably helpful when your functions require particular file system capabilities from Amazon FSx for NetApp ONTAP, Amazon FSx for OpenZFS, or Amazon FSx for Home windows File Server.
Pricing and availability
S3 Recordsdata is obtainable right this moment in all industrial AWS Areas.
You pay for the portion of information saved in your S3 file system, for small file learn and all write operations to the file system, and for S3 requests throughout knowledge synchronization between the file system and the S3 bucket. The Amazon S3 pricing web page has all the main points.
From discussions with clients, I consider S3 Recordsdata helps simplify cloud architectures by eliminating knowledge silos, synchronization complexity, and guide knowledge motion between objects and recordsdata. Whether or not you’re operating manufacturing instruments that already work with file methods, constructing agentic AI methods that depend on file-based Python libraries and shell scripts, or getting ready datasets for ML coaching, S3 Recordsdata lets these interactive, shared, hierarchical workloads entry S3 knowledge instantly with out selecting between the sturdiness of Amazon S3 and price advantages and a file system’s interactive capabilities. Now you can use Amazon S3 because the place for all of your organizations’ knowledge, figuring out the information is accessible instantly from any AWS compute occasion, container, and performance.
To be taught extra and get began, go to the S3 Recordsdata documentation.
I’d love to listen to how you utilize this new functionality. Be at liberty to share your suggestions within the feedback under.
— seb
