Launching S3 Files, making S3 buckets accessible as file systems | Amazon Web Services

Amazon S3 Files makes general-purpose S3 buckets accessible through a native NFS-based file system. It combines S3’s durability, cost, and broad service integration with interactive file operations, shared access, and low-latency performance. The post concludes that this reduces the need to choose between object storage and traditional file systems for many AWS workloads.

Bridging Object Storage and File Systems

• S3 Files presents S3 objects as files and directories.
• Applications can use standard NFS v4.1+ operations, including creating, reading, updating, and deleting files.
• Changes made through the file system are synchronized back to S3 as new objects or object versions.
• Changes made directly in S3 generally appear in the file system within seconds, though synchronization can sometimes take longer.
• Multiple compute resources can mount the same file system and share data without duplicating it.

Performance and Data Access

• S3 Files uses Amazon EFS underneath and provides approximately 1 ms latency for active data.
• Frequently accessed metadata and file contents are placed on high-performance storage.
• Large sequential reads can be served directly from S3 to maximize throughput.
• Byte-range reads transfer only the requested portion of a file, reducing data movement and cost.
• Intelligent prefetching anticipates access patterns.
• Administrators can choose whether to cache complete files or metadata only.
• NFS close-to-open consistency supports concurrent, interactive workloads such as ML pipelines and collaborative AI agents.

Supported AWS Compute Services

S3 Files can expose buckets to:

• Amazon EC2 instances
• Amazon ECS and EKS containers
• AWS Fargate workloads
• AWS Lambda functions

This allows production applications, machine-learning systems, and agentic AI tools to access shared S3 data using ordinary file-system interfaces.

Creating and Mounting an S3 File System

The demonstration uses an EC2 instance and a general-purpose S3 bucket:

• Create an S3 file system from the S3 console, AWS CLI, or infrastructure-as-code tools.

• Configure or discover a mount target inside the relevant VPC.

• Mount the file system on EC2 with commands such as:

  sudo mkdir /home/ec2-user/s3files
  sudo mount -t s3files fs-...:/ /home/ec2-user/s3files
  

• Files created in the mounted directory become visible in the S3 bucket after synchronization.

• Standard commands such as ls, echo, and AWS CLI operations can verify that file contents are consistent between the mount and S3.

Security, Permissions, and Monitoring

• IAM identity and resource policies control access at both the file-system and object levels.
• Data is encrypted in transit with TLS 1.3.
• Data at rest uses SSE-S3 or customer-managed AWS KMS keys.
• POSIX permissions rely on user IDs and group IDs stored as object metadata.
• CloudWatch provides performance and update metrics.
• CloudTrail records management events.
• EC2 instances should use the latest amazon-efs-utils package, which is included in AWS-provided AMIs.

S3 Files is best suited to workloads requiring shared, interactive file access while retaining data in S3. Teams should still evaluate access patterns and latency requirements, but the service offers a practical way to use familiar file operations without giving up S3’s centralized, durable storage model.