Thoroughly Tested
As the ONTAP AI architecture is thoroughly tested, deploying these solutions reduces challenges and the unknown.
Get in touch with our AI team.
NetApp AI Storage Solutions
Powered by ONTAP AI software, designed to manage, and protect your data
AI Storage by NetApp
ONTAP AI is a reference architecture design from NetApp combining AFF A-series storage arrays, NVIDIA DGX AI appliances linked by NVIDIA Networking switches - delivering the highest performance, the lowest latency and fastest throughput connectivity. These elements have been tested and optimised together and certified by NVIDIA to ensure a seamless AI training experience, offering a host of advantages.
Key Features
Accelerate Results
NetApp AFF A-series systems keep data flowing to GPUs with fast and flexible all-flash storage, featuring end-to-end NVMe technologies.
Scale Independently
The flexibility of the solution allows DGX compute and AFF storage to scale as required, independent of each other but with consistency.
Cost Effective
The range of AFF A-series storage solutions creates a range of capacity and cost entry points when designing and ONTAP AI reference architecture.
ONTAP AI Cluster Configurations
At the heart of the NetApp ONTAP AI architecture are NVIDIA DGX H100 systems, supporting the highest performing training and inferencing a single platform. Each DGX H100 system is powered by eight NVIDIA H100 Tensor Core GPUs and features high-speed NVIDIA ConnectX-7 interconnects. These link to NetApp AFF A-series storage arrays, capable of feeding data to the DGX systems up to 4 times faster than competing solutions. The solution is completed with NVIDIA Spectrum Ethernet switches, which provide the low latency, high density, high performance, and the power efficiency demanded by AI environments.
The NetApp AFF Storage Array Range
ONTAP AI is a reference architecture design from NetApp combining AFF A-series storage arrays, NVIDIA DGX H100 Universal AI appliances and NVIDIA Mellanox Ethernet switches for fast connectivity. These elements have been tested and optimised together ensuring a seamless AI training experience, and offers a host of advantages including the following.
- A massive throughput of up to 300Gbps and 11.4 million IOPS in a 24-node cluster
- 30TB solid-state drives (SSDs) with multi-stream write (MSW)
- 100Gb Ethernet together with 32Gb FC connectivity
- High density with 2PB in a 2U drive shelf
- Scaling from 364TB (2 nodes) to 74PB (24 nodes)
- ONTAP 9.8, with a complete suite of data protection and replication features
AFF A-Series systems, from entry-level to high- end, support end-to-end NVMe technologies, from NVMe-attached SSDs to front-end NVMe over Fibre Channel (NVMe/FC) host connectivity. These systems deliver the industry’s lowest latency for an enterprise all-flash array, making them a superior choice for driving the most demanding workloads and applications. With a simple software upgrade to the modern NVMe/FC SAN infrastructure, you can drive more workloads with faster response times, without disruption or data migration.
 |
 |
 |
 |
|
| AFF A900 | AFF A800 | AFF A400 | AFF A250 | |
|---|---|---|---|---|
| Best For: | The most demanding workloads requiring ultra-low latency | Performance-driven workloads and mission-critical applications requiring high resiliency | Most enterprise applications that require best balance of performance and cost | Mid-size business and small enterprises that require simplicity and best value |
| Maximum Scale Out | 24 nodes | 24 nodes | 24 nodes | 24 nodes |
| Maximum SSD | 5760 | 2880 | 5760 | 576 |
| Maximum Effective Capacity | 702.7PB | 316.3PB | 702.7PB | 35PB |
| Software | ONTAP Enterprise | ONTAP | ONTAP | ONTAP |
| Controller Chassis Form Factor | 8U | 4U | 4U | 2U |