NetApp has refreshed its AFF all-flash storage arrays amid a marketing focus on artificial intelligence (AI) workloads and cutting energy waste in the datacentre.
Alongside this comes an avowed target of more customers for its block access arrays and the availability of QLC drives across its Storagegrid object storage family, plus some additions in its Ontap OS and product bundles for Nvidia-based workloads.
Key among the announcements are refreshes to the all-flash AFF family, which include the AFF A70, which replaces the mid-range A400; the A1K, which goes for the high end where the A900 sat previously; and the A90, which sits at the high to mid end and replaces the A700.
According to Grant Caley, NetApp director for solutions engineering, the latest hardware represents “unified storage for the AI era”. Here, said Caley, unified means “any data type, any workload, any protocol, anywhere”, with security built in.
Capacities stretch to hundreds of petabytes (PB) for clusters of the high-end product, with up to 30TB (terabytes) QLC drives available for capacity configurations and 15TB TLC on performance-focused arrays.
While a key focus of the marketing around the launch is the increased prevalence of AI workloads, Caley said NetApp had a much wider target market, which included VMware deployments and databases.
“They’re designed for any workload, not a specific AI workload profile,” he said, although he did highlight NetApp functionality that allows data scientists to prepare large datasets for AI, such as being able to make clones and copies quickly.
Caley also made a point of targeting block workloads, simply, he said, because NetApp – long known as a NAS “filer” specialist – hasn’t fulfilled it’s potential in that part of the market. “We’ve got 20,000 customers that run our arrays as SAN but it’s such a small part of the market and we can get more. It’s an opportunity for NetApp,” he said.
For the new AFF arrays, Caley claimed big performance upticks on the previous generation, namely 50% lower $/IOPS, 55% lower $/GBps (throughput), 35% lower $/effective TB, and 45% lower $/density.
Core to the array launch is a set of better-performing arrays, but NetApp is keen to wrap these in a message in which it calls itself an “intelligent data infrastructure company”.
Caley said: “The intelligent part is the ability to move data from one box to another, from one location to another. Also tiering of hot to colder tiers, to an S3 object store or on-prem high-performance to [Microsoft] Azure or GCP [Google Cloud Platform] cold storage, for example.”
According to Matt Watts, chief technology evangelist at NetApp, this comprises a “unified storage” core that offers a single operating environment across on-premise and cloud – where NetApp is the only storage company with its products available in all three hyperscalers – plus intelligent data services with security and anti-ransomware functionality built into storage, and a focus on sustainability.
Watts said currently 2% of worldwide electricity use is in datacentres, with 8% projected by 2030, and 38% of that being consumed by storage. At the same time, he said, most data (68%) is never accessed more than once and sustainability should largely comprise targeting waste in the datacentre.
“Some vendors focus on hardware,” said Watts. “But we think we should find waste in places like unused VMware licences and CPU use, and the 68% of data that never gets used again, and decide what to do with it. Then you can look at the datacentre and use of systems that are not power-efficient.”
While the new arrays were the focus of the NetApp launch, the company also announced a collaboration with Lenovo to provide what it called “generative AI in a box”.
This is the AIPod with Lenovo for Nvidia OVX. It comes soon after NetApp announced a number of other products that link with hardware from the GPU and AI infrastructure market leader. These were AIPod with Nvidia DGX, AIPod with Nvidia OVX, Ontap with Nemo Retriever, and FlexPod for AI.
Finally, NetApp also announced that its Ontap operating system has moved to version 9.15.1, while its Storagegrid object storage family now comprises five new models, defined by their upgrade to high-density QLC flash drives.