Estimate AWS FSx costs for Lustre, Windows File Server, and NetApp ONTAP.
Last verified: May 2026
Lustre throughput is included in the storage price.
Output will appear here...The estimator computes monthly FSx cost across the four file system types: Lustre (scratch GB × rate, persistent GB × rate, optional throughput capacity), Windows (SSD storage × tier rate, throughput, AD integration cost, Multi-AZ multiplier), ONTAP (SSD GB + capacity pool GB + throughput capacity + backup), OpenZFS (SSD GB + throughput + snapshot retention). Output presents component-level breakdown plus comparison against equivalent EBS+EC2 self-managed file server cost.
Amazon FSx provides fully managed file systems optimized for different workloads — FSx for Lustre delivers high-performance parallel storage for HPC and ML, FSx for Windows File Server offers native SMB protocol support with Active Directory integration, FSx for NetApp ONTAP provides multi-protocol NAS with data management features, and FSx for OpenZFS offers high-performance ZFS-based storage. Costs vary significantly by file system type, storage capacity, throughput, IOPS provisioning, and backup retention. This estimator helps you calculate monthly costs across all FSx variants with region-specific pricing.
Your team's ML training pipeline reads 5 TB of training data from a self-hosted Lustre cluster on EC2 (4 instances, $2,400/month). The estimator models FSx Lustre scratch alternative: 5 TB × $0.14/GB-mo + 1.2 GB/s throughput included = $716/month. Setup eliminates the operational burden of managing the Lustre cluster (kernel updates, monitoring, scaling). Net savings: $1,684/month + ~10 hours/month of engineer time.
FSx for Lustre scratch is dramatically cheaper than persistent — but only if you can actually reload data from S3 in case of failure. For ML training where data is in S3 anyway, scratch is the right choice. For data that's expensive to recreate (intermediate state, processing pipelines), pay for persistent.
FSx for ONTAP's auto-tiering between SSD and capacity pool is the killer cost feature. Hot data (recently accessed) stays on fast SSD; cold data automatically moves to capacity pool at ~80% lower cost. For mixed-access workloads (DB + cold archive), this can cut storage costs by 50%+ without app changes.
Multi-AZ FSx for Windows costs roughly 2x Single-AZ — but is the only option if you need uptime SLA during AZ events. For dev/test, Single-AZ is fine. For production file shares serving Active Directory or critical apps, Multi-AZ is justified. For DR-only file servers, consider cross-region snapshot replication instead of Multi-AZ to control costs.
Scratch file systems are lower cost because data is not replicated — if a server fails, data must be reloaded from S3 or regenerated. Persistent file systems replicate data within the same AZ and cost more per GB-month. Scratch is ideal for temporary processing (ML training, HPC simulations) where data can be reloaded. Persistent is required for workloads that cannot tolerate data loss. Both types include baseline throughput with the storage, and you can provision additional throughput capacity for extra cost.
ONTAP costs have three main components: SSD storage (high-performance tier), capacity pool storage (lower-cost tier for infrequently accessed data), and throughput capacity. Data automatically tiers between SSD and capacity pool based on access patterns, which can significantly reduce costs. You also pay for backups and any inter-AZ data transfer in Multi-AZ deployments. Data deduplication and compression (included at no extra cost) can reduce the effective storage consumed by 30-65% depending on your data.
Was this tool helpful?
Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.