Amazon FSx: Managed File Systems for Windows, Lustre, NetApp, and OpenZFS Workloads

EFS handles Linux NFS workloads. But some applications need something more specific — Windows file shares that speak SMB, high-performance parallel file systems for supercomputing, or enterprise storage appliances that the dev team already knows. Amazon FSx is AWS’s answer: a family of fully managed file systems based on industry-standard technologies, each managed by AWS at the infrastructure level but identical in behavior to the systems your team already uses on-premises.

There are four FSx products, each targeting a distinct category of workload.

FSx for Windows File Server

FSx for Windows File Server runs on actual Windows Server infrastructure managed by AWS. It exposes SMB shares that Windows EC2 instances, on-premises Windows servers, and Windows 10/11 workstations can connect to with standard Windows Explorer or net use commands. Because it is built on Windows Server, it supports:

Active Directory integration — joins your existing AD domain, enforcing the same users, groups, and permissions as your on-premises environment
DFS Namespaces — present multiple file shares as a unified namespace to users
Shadow Copies (VSS) — point-in-time snapshots accessible through Windows Previous Versions
SMB 3.0 with continuous availability failover

  FSx for Windows File Server — Architecture
  ==========================================

  On-Premises                        AWS Region
  ─────────────────────────────────────────────────

  Active Directory ────────────────► FSx joins the domain
  Domain Controller                  │
                                     │
  Windows Workstations               FSx for Windows File Server
  (\\fileserver\share)  ──────────► (Multi-AZ, SMB 3.0)
                                     │
  EC2 Windows Instances ───────────► Same SMB share

  Users see the same permissions, paths, and file structure
  regardless of whether they connect from on-premises or EC2

When to use it: migrating Windows file servers to the cloud without changing how users connect, running SQL Server backup files on SMB, or providing home directories for Windows users in a hybrid environment.

FSx for Lustre

Lustre is a parallel, distributed file system used in supercomputing and high-performance computing. It is the file system behind most of the world’s fastest supercomputers. FSx for Lustre brings that same technology to AWS, managed by Amazon, with native integration with S3.

Key characteristics:

Sub-millisecond latency for individual file operations
Hundreds of GB/s aggregate throughput by distributing data across multiple storage servers
S3 integration: import data from S3 at file system creation; changes can be exported back
Scratch and Persistent deployment types: scratch for temporary processing, persistent for long-running jobs that need data to survive hardware failures

  FSx for Lustre — HPC Data Flow
  ================================

  S3 Bucket ──────────────────────────────────────────────►
  (source data: 50 TB genome sequences)                    │
                                               FSx Lustre  │
  ┌─────────────────────────────────────────────────────┐  │
  │  Compute nodes (EC2 C5n, 200 nodes)                 │◄─┘
  │  Each node mounts FSx Lustre                        │
  │  All nodes read/write same files concurrently       │
  │  Aggregate throughput: 100+ GB/s                    │
  └─────────────────────────────────────────────────────┘
                │
                ▼
         Results written back to S3

  Duration: hours instead of days on traditional storage

When to use it: genomics analysis, seismic data processing, financial Monte Carlo simulations, machine learning training with large datasets that need fast random access, video rendering pipelines.

FSx for NetApp ONTAP

NetApp ONTAP is the storage operating system behind NetApp’s enterprise storage appliances — widely used in data centers for NFS, SMB, and iSCSI workloads. FSx for NetApp ONTAP runs a fully managed version in AWS that is operationally identical to an on-premises NetApp array.

Key capabilities:

Multi-protocol: serves NFS (Linux), SMB (Windows), and iSCSI simultaneously from the same file system
SnapMirror replication: replicate data to another ONTAP system on-premises or in another AWS region
Storage efficiency: deduplication, compression, and thin provisioning reduce actual storage consumption
FlexClone: create instant, writable clones of volumes or files without copying data — invaluable for dev/test environments

When to use it: lift-and-shift of existing NetApp workloads, multi-protocol environments where Linux and Windows servers share the same data, organizations that already use NetApp APIs and tooling.

FSx for OpenZFS

OpenZFS is an open-source file system with a strong reputation for data integrity, compression, and snapshots. FSx for OpenZFS provides fully managed OpenZFS file systems accessible over NFS.

Key capabilities:

Data integrity: ZFS uses checksums to detect and correct silent data corruption
Instant snapshots: space-efficient snapshots with no performance impact
Compression and deduplication: built-in, configurable per dataset
Up to 1 million IOPS with sub-millisecond latency

When to use it: Linux workloads migrating from on-premises ZFS, applications that need strong data integrity guarantees, databases and analytics that benefit from compression reducing I/O.

Choosing the Right FSx Product

  FSx Product Decision Tree
  ==========================

  What is the primary protocol?
  ├── SMB (Windows file shares, AD)  ──────────► FSx for Windows File Server
  ├── NFS for HPC / parallel I/O    ──────────► FSx for Lustre
  ├── Multi-protocol (NFS + SMB + iSCSI) ─────► FSx for NetApp ONTAP
  └── NFS with strong data integrity / ZFS ───► FSx for OpenZFS

  Existing on-premises dependency?
  ├── NetApp arrays                  ──────────► FSx for NetApp ONTAP
  ├── Windows file servers + AD      ──────────► FSx for Windows File Server
  └── ZFS-based systems              ──────────► FSx for OpenZFS

Real-World Use Case: Game Studio Rendering Pipeline

A game studio runs a rendering farm of 500 EC2 spot instances that generate frames for cinematic cutscenes. Each frame requires reading shared texture assets (5 TB total) and writing completed frames. The workload is:

Highly parallel — all 500 instances access the same textures simultaneously
Throughput-bound — each instance reads hundreds of MB per second
Time-limited — a render job runs for 4-8 hours then the fleet terminates

FSx for Lustre in scratch mode is the right tool: import texture assets from S3 at job start, run the render across all instances with aggregate throughput in the hundreds of GB/s range, then export completed frames back to S3. The file system is deleted when the job ends, and the studio pays only for the hours it ran.

Key Interview Points

FSx for Windows requires joining an Active Directory domain — it will not work standalone without AD
Lustre scratch vs persistent: scratch does not replicate data between storage servers; persistent does. Use scratch for temporary jobs, persistent for production HPC pipelines
FSx for NetApp ONTAP is the only FSx product that supports iSCSI, enabling block-level access alongside file-level access from the same system
FSx is not EFS — EFS is AWS-native NFS; FSx is managed third-party file systems. Different pricing models, different protocols, different use cases
Multi-AZ availability differs by product: FSx for Windows and NetApp ONTAP support Multi-AZ; FSx for Lustre and OpenZFS are single-AZ
S3 integration is unique to Lustre among the FSx products — the others do not have direct S3 data repository linking