The News: Amazon Web Services (AWS) enhanced its object and file system storage services at AWS re:Invent, which fit into its overall theme of improving its public cloud for enterprises looking to use generative AI. You can find storage announcements in blog posts on the AWS website.

AWS re:Invent: With an Eye on AI, AWS Adds, Enhances Storage Services

Analyst Take: As usual, AWS made a wide range of storage and data protection announcements at re:Invent this year. Here is the main storage news.

Object Storage

Amazon S3 Express One Zone. AWS claims this new storage class can deliver up to 10 times improved performance on smaller objects than S3 standard storage. It also claims it can handle hundreds of thousands of requests per second with single-digit millisecond latency. The objects are stored and replicated – likely on flash storage – within one AWS Availability Zone to reduce the latency between compute and storage. The Express One Zone class is especially beneficial for smaller objects because there is less data to read.

File Storage

Elastic File System (EFS) Archive. Archive is a new storage class for the AWS EFS for running file workloads. EFS Archive keeps coldest file data always available. At $0.008/GB per month, EFS Archive costs up to 97% less than EFS Standard and up to 50% lower than EFS Infrequent Access storage classes in the US East region. EFS Archive is for file data accessed no more than a few times a year. Intelligent tiering can automatically move files from EFS Standard with sub-millisecond SSD latencies to EFS Infrequent Access to EFS Archive based on the last time the files were accessed.

FSx for NetApp ONTAP. AWS also enhanced its enterprise-class file system co-engineered with NetApp. Enhancements include new scale-out file systems, Virtual Private Cloud (VPC) support, and FlexGroup volume management.

Until now, FSx for ONTAP has been only scale-up, running on a single pair of servers in active-passive high-availability configurations. The new scale-out FSx for ONTAP file systems option uses from two to six high-availability pairs. While the scale-up file systems support a maximum of 4 GBps of read throughput, 1.8 GBps of write throughput, 160,000 IOPs, and 192 TiB of SSD storage, the scale-out file systems support up to 36 GBps of read throughput, 6.6 GBps of write throughput, 1.2 million IOPs, and 1 TPiB of SSD storage.

Customers who specify 4 GBps or less throughput will get the scale-up server configuration while customers with more than 4 GBps throughput receive scale-out server configurations. Scale-out file systems can use multiple availability zones while scale-out systems are only available in single AZs.

FSx for ONTAP users can now create multi-AZ file systems in VPCs shared with them by other accounts. That makes it possible to create and access high-available storage from multiple VCP virtual networks.

Users can also create, manage, and back up FSX for ONTAP FlexGroup volumes through the AWS Management Console, the Amazon FSx CLI, and the AWS SDK. Previously, they could only create FlexGroups using the ONTAP CLIP and ONTAP REST API. Customers can also now create Amazon FSx backups of r FlexGroup volumes. A FlexGroup volume is a scale-out NAS container that uses automatic load distribution and scalability for high performance. A FlexGroup can scale to 20 PBs.

On-demand replication for FSX for OpenZFS – this feature enables customers to send a snapshot from one file system to another file system in their FSx for OpenZFS account. FSx for OpenZFS file systems are accessible from Linux, Windows, and macOS compute instances and containers through the NFS file protocol.

Impact on AI

What does all this have to do with AI? Well, object and file storage are used mainly for unstructured data, such as audio, video, image, and document content frequently utilized for AI. Amazon S3 Express One Zone keeps storage and compute closer together, and the EFS Archive makes cold data more available. Placing compute and storage close is important to generative AI because it keeps data closer to training nodes. It also speeds monitoring and inferencing for AI.

These enhancements—along with the general performance gains—can be of great benefit to customers who have petabytes of data in AWS that they want to use for generative AI.

Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.

Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.