Deployment and storage class options for FSx for Lustre file systems
FSx for Lustre file systems use the persistent deployment type.
Persistent file systems
Persistent file systems are designed for longer-term storage and workloads. The file servers are highly available and data is automatically replicated within the same Availability Zone in which the file system is located. The data volumes attached to the file servers are replicated independently from the file servers to which they are attached.
Amazon FSx continuously monitors persistent file systems for hardware failures, and automatically replaces infrastructure components in the event of a failure. On a persistent file system, if a file server becomes unavailable, it's replaced automatically within minutes of failure. During that time, client requests for data on that server transparently retry and eventually succeed after the file server is replaced. Data on persistent file systems is replicated on disks, and any failed disks are automatically replaced transparently.
Use persistent file systems for longer-term storage and for throughput-focused workloads that run for extended periods or indefinitely, and that might be sensitive to disruptions in availability.
Persistent deployment types automatically encrypt data in transit when they are accessed from Amazon EC2 instances that support encryption in transit.
Persistent deployment types can be built on Lustre v2.10 or v2.12, and support SSD (solid state drive) and HDD (hard disk drive) storage types. For SSD storage, the throughput per unit of storage is either 50, 100, or 200 MBps per tebibyte (TiB). For HDD, throughput per unit of storage is 12 or 40 MBps per tebibyte.
Persistent 1 deployment types are available in all Amazon Web Services Regions.
IP addresses for file systems
Each FSx for Lustre file system requires one IP address for each metadata server (MDS) and one IP address for each storage server (OSS).
| File System Type | Throughput, MBps/TiB | Storage per OSS |
|---|---|---|
| Persistent 1 SSD | 50, 100, 200 | 2.4 TiB per OSS |
| Persistent HDD | 12 | 6 TiB per OSS |
| 40 | 1.8 TiB per OSS |
Note
* Amazon FSx provisions a metadata server for every 12,000 Metadata IOPS on Persistent 2 SSD and Intelligent-Tiering file systems configured with metadata configuration.
Amazon FSx for Lustre Intelligent-Tiering file systems support a maximum of 512 TiB of storage per OSS.
FSx for Lustre storage classes
Amazon FSx for Lustre offers solid state drive (SSD) and hard disk drive (HDD) storage classes that are optimized for different data processing requirements:
The SSD storage class provides low-latency (sub-millisecond) access to your full dataset. The SSD storage class is provisioned, which means that you specify a file system size and pay storage costs for the amount of storage provisioned. Use the SSD storage class for latency-sensitive workloads that require the performance of all-flash storage across all data.
Persistent 2 file systems with SSD storage support higher levels of throughput per unit of storage (that is, 250, 500, or 1000 MBps per TiB) compared to Persistent 1 file systems. For a Persistent 1 file system with SSD storage, the throughput per unit of storage is either 50, 100, or 200 MBps per TiB. For a Scratch file system with SSD storage, the throughput per unit of storage is 200 MBps per TiB.
The HDD storage class can be used with workloads that need consistent single-digit ms latency across all data. You can provision an optional SSD read cache that is sized to 20% of your HDD storage capacity to provide low-latency access to frequently-accessed data. With HDD storage, you specify a file system size and pay for the amount of storage that you provision. For a Persistent 1 file system with HDD storage, the throughput per unit of storage is either 12 or 40 MBps per TiB.
For more information about performance of these storage classes, see Performance characteristics of SSD and HDD storage classes and .
How the Intelligent-Tiering storage class tiers data
The Amazon FSx Intelligent-Tiering storage class automatically stores data in three access tiers. It is designed to optimize storage costs by automatically moving data to the most cost-effective access tier, without performance impact or operational overhead. The Intelligent-Tiering storage class automatically tiers data based on last access time, thus automatically optimizing costs for less active data:
Data accessed in the last 30 days is stored in the Frequent Access tier.
Data that hasn’t been accessed in 30 consecutive days automatically moves to the Infrequent Access tier, and costs less than data in the Frequent Access tier.
Data that hasn’t been accessed in 90 consecutive days automatically moves to the Archive Instant Access tier, and costs less than data in the Infrequent Access tier.
When you access data in the Infrequent Access or Archive Instant Access tiers, the data automatically moves back to the Frequent Access tier. All access to non-cached data has the same performance characteristics, independent of the data’s tier, and there are no additional IOPS, retrieval, or transition costs beyond your normal read/write operation costs.