Amazon FSx
Amazon FSx is the Amazon Web Services family of fully managed file systems, letting customers launch and scale a file system built on an established, familiar file-system technology rather than operating that software themselves. It spans four member services — Amazon FSx for Windows File Server, Amazon FSx for Lustre, FSx for NetApp ONTAP, and FSx for OpenZFS — each suited to a different workload, and sits alongside Amazon Elastic File System (EFS) as an alternative when a workload needs a specific file system’s protocol or feature set rather than EFS’s Linux/NFS-only, fully elastic model.
Key points
- Amazon FSx for Windows File Server — SMB-based shared storage built on Windows Server, for Windows-native applications, home directories, and workloads needing Microsoft Active Directory-integrated identity.
- Amazon FSx for Lustre — a parallel, high-performance file system for machine learning, analytics, and HPC workloads that need sub-millisecond latencies and up to millions of IOPS.
- FSx for NetApp ONTAP — fully managed shared storage built on NetApp’s ONTAP file system, aimed at migrating legacy NetApp/NAS workloads and hybrid-cloud scenarios that rely on ONTAP data-management and replication features.
- FSx for OpenZFS — fully managed storage built on the open-source OpenZFS file system, for teams already standardized on ZFS’s snapshot, cloning, and compression capabilities.
- Choosing among the four is typically driven by which underlying file system a team already knows, or by matching each system’s protocol, performance profile, and data-management features to the workload’s requirements — not by a single one-size-fits-all default.
- Distinct from Amazon Elastic File System (EFS): EFS is AWS’s own NFS-based, serverless, fully elastic Linux file storage with no underlying file-system choice to make; the FSx family exists specifically for workloads that need Windows/SMB, Lustre, ONTAP, or OpenZFS semantics that EFS does not provide.
- All four FSx file systems are fully managed — AWS handles patching, hardware provisioning, and updates to the underlying file servers, not the customer.
- Amazon FSx for Lustre S3 integration (one of the most-tested FSx facts on the exam): FSx for Lustre can link an Amazon S3 bucket as a data repository association, lazily loading S3 objects as files on first access and optionally exporting processed results back to the linked bucket — making it well suited to HPC and machine-learning training workloads that read training data from, and write results back to, S3.