Accelerate file workload performance with second-generation Amazon FSx for NetApp ONTAP file systems

Supporting file-based workloads with growing performance demands over time is a continuous challenge for organizations. As datasets expand, static infrastructure struggles to keep pace, resulting in disruptive migrations to new infrastructure. Organizations need highly scalable file storage that delivers speed at-scale today, while adapting seamlessly to future requirements.

Amazon FSx for NetApp ONTAP provides fully managed ONTAP file systems natively in Amazon Web Services (AWS). It is purpose-built for fluctuating workloads, allowing you to scale performance and storage without disruption. With FSx for ONTAP’s second-generation file systems, we are thrilled to announce enhancements that allow you to scale performance up to 18 times higher than with first-generation file systems.

In this post, we review second-generation file systems in more detail. We start with a background on FSx for ONTAP’s performance scalability. Then, we introduce the enhanced performance scalability and flexibility second-generation file systems offer over previous-generation file systems, and showcase two new features offered by second-generation file systems, 1/the ability to more quickly access your files from a data backup, and 2/the ability to use the NVMe-over-TCP protocol as a modernized, simplified, and faster alternative to iSCSI block storage. Finally, we show you how to create and update second-generation file systems.

Background on Amazon FSx for NetApp ONTAP performance and scalability

Before today, FSx for ONTAP file systems came in two varieties, 1/scale-up and 2/scale-out file systems. Scale-up file systems supported up to 4 GBps of throughput and 192 TiB of provisioned solid-state drive (SSD) storage on a single highly-available (HA) pair of file servers. Scale-out file systems offered the performance of multiple file systems in one by automatically spreading the workload across up to 12 HA pairs of file servers—each with up to 6 GBps of throughput—allowing you to scale up to 72 GBps of throughput and up to 1 PiB of provisioned SSD storage.

Scale-up file systems fit the needs of most of today’s workloads, e.g., general-purpose file shares, applications, and databases. However, a workload’s performance requirements tend to increase as it grows in size and as customers do more with their data. Larger, more demanding workloads are well-suited for scale-out file systems, but before today you’d need to migrate your data to a new file system if your performance demands exceeded what your existing file system could offer. This is because you couldn’t add HA pairs or change the throughput capacity for your file system’s existing HA pairs.

Introducing second-generation file systems

Second-gen file systems allow you to create file systems with a single HA pair that delivers up to 6 GBps of throughput (a 50 percent increase from first-generation) and 512 TiB of SSD storage (a 160 percent increase from first-generation), giving your workloads even more room to grow within a single HA pair. If you ever need more performance than your existing HA pairs offer, you can add new HA pairs to single availability zone(AZ) file systems. You can also change the throughput of file systems with multiple HA pairs in just a few choices, making it easier than ever for your file system to keep pace with your workloads.

With support for up to 72 GBps aggregate throughput across a multi-pair second-gen file system, FSx for ONTAP can meet the needs of even the most demanding workloads, such as electronic design automation (EDA), visual effects (VFX) rendering, machine learning training pipelines, and petabyte-scale databases. Plus, with 50 percent higher baseline throughput per HA pair—from 4 GBps before up to 6 GBps now—and up to 300 percent higher burst capabilities (6.2 GBps vs. 1.5 GBps network throughput burst, and 3.1 GBps vs. 1.2 GBps disk throughput burst), your dynamic workloads have ample headroom to smoothly handle temporary spikes.

The following specification table compare first-generation and second-generation file systems:

Speed up file recovery times with near-instantly accessible backup restores

With second-generation file systems, FSx for ONTAP debuts a new capability that reduces backup recovery times for your files. Traditionally when restoring FSx for ONTAP volume backups, you had to wait for your entire dataset to be restored before you could access any data within it. Now, when you restore a backup on a second-gen file system, you get read access to the volume within minutes of initiating the restore, allowing you to read your entire backup data set up to 17x faster compared to backup restores on first-generation file systems.

This enhanced restore capability unlocks the ability to rapidly retrieve critical files after accidental deletion without waiting on full restore. For example, you can now restore a single file or entire directories without waiting for the entire volume to be restored by initiating a restore, transferring the data you need off of it, and then canceling the restore—even before it has completed.

Accelerate block storage with the NVMe-over-TCP protocol

Second-generation file systems also offer a new block storage protocol “NVMe-over-TCP”. An alternative to iSCSI block storage, NVMe-over-TCP provides a modernized and simplified experience. For example, with iSCSI you need to manage multipathing between file system nodes which allows your clients to failover between your file system’s file servers, but with NVMe-over-TCP, it’s included within the protocol. And, its more efficient networking stack can offer lower latencies for some workloads, improving performance over iSCSI.

Creating second-generation file systems

You can create a second-gen FSx for ONTAP file system using the AWS Management Console, the AWS Command Line Interface (AWS CLI), or by writing code that calls the Amazon FSx CreateFileSystem function. Using the Console as an example, start by choosing Amazon FSx for NetApp ONTAP as shown in Figure: 1.

Figure: 1 – Select file system type

You can create second-generation file systems either through Quick create or Standard create as shown in Figure: 2. For Quick create, second-gen file systems are the default file system type in all Regions where they’re available. For Standard create, you can choose either second-gen or previous-gen deployments.

To keep things simple, choose Quick create. Then, enter a name, select Single-AZ deployment, and enter your desired SSD storage capacity. You can accept the recommended throughput capacity (based on your SSD storage) or enter your desired throughput value:

Figure: 2 – Creation method

When you choose to specify your own throughput, you’re presented with a list of available options that most closely match your desired throughput a shown in Figure: 3.

Figure: 3 – Throughput capacity options

For some desired values, the Console will show you multiple total throughput options that consist of a different number of HA pairs. The following are guidelines to help you choose a configuration for your workload:

• Lower throughput – A single HA pair offers the lowest throughput capacities (384 MBps, 768 MBps) and lowest minimum SSD storage capacity (1 TiB). If you expect your workload to only need up to 6 GBps of throughput over its lifetime, then you can cost-optimize by choosing a single HA pair.
• Higher throughput – Total throughput and SSD storage scalability increase with the number of HA pairs your file system has. While you can always add HA pairs to second-gen file systems, starting with the number of HA pairs that allow you to scale throughput to the maximum you expect your workload to need over its lifetime simplifies scaling your workload over time. You need to move your data around your file system after adding HA pairs, but you don’t when scaling throughput and storage for your existing HA pairs. For example, if your workload only needs 6 GBps of throughput today but you expect it’ll need as much as 24 GBps in the future, then choosing a 4-pair configuration allows you to start with 6,144 MBps of throughput (1,536 MBps per HA pair), and later scale up to 12,288 MBps (3,072 MBps per HA pair) then 24,576 MBps (6,144 MBps per HA pair) in response to your workload’s growth.
• Block storage – ONTAP supports the iSCSI and NVMe-over-TCP protocol for file systems with up to 6 HA pairs. If you plan to base your workload on one of these protocols, then choose a file system configuration with up to 6 HA pairs. Keep in mind that you can’t expand your file system beyond 6 HA pairs without losing access to iSCSI and NVMe-over-TCP.

In Quick create, you also choose your file system’s virtual private cloud (VPC) and whether storage efficiency is enabled for the first volume that’s created for you. After making your selections, select Next to review your settings, then click Create file system. It typically takes about 20 minutes for FSx to create your file system.

Updating performance and storage over time

After your file system has been created, you can update your throughput, storage, provisioned IOPS, and—for Single-AZ file systems—HA pairs using the Console, CLI, or SDK. Using the Console as an example, simply select your file system and you’ll see an option to update each as indicated in Figure: 4.

Figure: 4 – Updating performance

Cleaning up

After you’ve created your file system, you can clean up any resources to avoid unnecessary charges by following the steps in the Cleaning up resources section of the FSx for ONTAP user guide.

Summary

Second-generation file systems offer up to 18 times higher performance scalability than first-generation file systems. Not only can you scale the throughput of each HA pair of your file system up to 6 GBps, you can also create or expand your file system with up to 12 HA pairs. And, you get the benefit of faster data access when restoring from backups, and a simplified alternative to iSCSI for your block storage workloads via the NVMe-over-TCP protocol. These enhancements open the door to supporting more demanding workloads than ever before. Second-gen file systems are available today in the following AWS Regions:US East (N. Virginia, Ohio), US West (Oregon), Europe (Ireland), and Asia-Pacific (Sydney). To learn more, visit the FSx for ONTAP user guide.