NetApp maintains push to data management for AI

0
77
VMware CEO tells enterprises to become 'cloud-smart' to speed up pace of digital transformation

Source is ComputerWeekly.com

NetApp will maintain its big push towards data management for artificial intelligence (AI) workloads and took the opportunity to explain that move at its Insight 2024 event in Las Vegas this week.

Core to that is data management via a global metadata namespace in its Ontap storage operating system that aims to make corporate data easier to find, classify, explore, process and store data for AI modelling and inference.

Also announced were new block storage ASA and file-oriented FAS arrays, additions to ransomware protection in the Ontap operating system and BlueXP control plane, and new NetApp storage services in Microsoft Azure and Google Cloud. 

Data management for AI

NetApp set out its vision for AI innovation in a way that nudges it further into data management for AI territory.

Core to its vision is a set of services that allow data to be gathered together via custom criteria, whether on-premise or in the cloud and in any location. This can be done via natural language questioning and resulting collections made available for AI processing.

NetApp CEO George Kurian said: “We’re at an interesting point in the evolution of the business and technology landscape, the junction of data and intelligence.

“There’s a gap between what the AI systems can do – the chips, the systems – and the data systems. NetApp aims to bring intelligence to the infrastructure, and the data on the infrastructure, to provide a set of tools to understand and explore the data so it can be selected and used.”

More specifically with reference to the mechanism of AI, Kurian spoke of the need to bring more intelligence to handling models, to model traceability and versioning and security of data over its lifecycle with policies that travel with the data.

Krish Vitaldevera, NetApp’s senior vice-president and general manager for platform, expanded on the topic. “It all comes down to that we think AI is a data problem, but it is complex and costly,” he said.

Vitaldevera identified in particular vector database bloat, where, for example, during retrieval-augmented generation (RAG) pipeline use cases, the number of dimensions in the data are multiplied, whitespace is added, more data is added, which can all see total data expand by eight or 10 times.

According to Vitaldevera, Ontap’s new functionality aims to resolve that through data reduction. 

New ASA block and FAS filers

Meanwhile, NetApp also announced addition to its ASA and FAS storage array families. 

NetApp’s senior vice-president and general manager for enterprise storage, Sandeep Singh, said: “What we’re trying to solve is that customers often face a choice between simple block storage and storage that has scale-out capabilities and advanced data management. What if it’s possible to get complexity and power without the trade-off?”

With that in mind, NetApp has launched the ASA A70, A90 and A1K block storage arrays, which it claims have gained a performance boost of 109% (A70), 88% (A90) and 50% (A1K) over existing versions. 

It also claims lower upfront costs than its competitors, by between 20% and 35%. “It puts modernisation within every budget,” said Singh, adding that workloads targeted are VMware, Oracle, SAP, SQL databases and Epic healthcare record systems.

Meanwhile, NetApp has announced new FAS arrays aimed at secondary workloads, such as backup. Here, the new FAS70 claims an 85% speed boost over the FAS8300 and the new FAS90 35% over the existing FAS9500.

The new ASA A series adds a block-optimised storage family to the existing AFF file-oriented A series performance flash series launched in May. The new FAS arrays offer a lower tier of hybrid flash (with HDD) with capacity-optimised quad-level cell (QLC) flash-equipped AFF and ASA C-series in between. 

Ransomware upgrades

NetApp announced upgrades to ransomware protection and response in Ontap and its BlueXP control plane.

In Ontap – now up to version 9.16.1 – its autonomous ransomware protection has reached general availability (GA). This amounts to automated ransomware profile updates, AI-based detection with a claimed 99% precision, and creation of immutable snapshots in response.

BlueXP adds a dashboard where customers can scan their data for potential risks – such as the presence of personally identifiable information (PII) – categorise workloads and prioritise levels of protection. It also provides the ability to create alerts to suspicious user activity (available at the end of 2024).

Integration is also possible that allows ransomware alerts to be sent directly to an organisation’s security ecosystem to accelerate the detection-response cycle. 

Cloud options boosted

In Azure NetApp Files, where NetApp storage is available as a service, a new Cool Access option will be available. This tiers data to cheaper Azure Blob (object) storage when use becomes less frequent but restores it when needed by the app or user. 

In Google Cloud NetApp Volumes, its value Flex option is now available in 40 Google Cloud regions. Also in Google Cloud, auto-tiering moves data from higher service tiers to lower-cost storage.

VMware audit capability

Concern over VMware costs is a current trend following licensing changes after the virtualisation giant’s purchase by Broadcom. Here, NetApp now offers an automated service from within BlueXP that provides an audit of the customer’s VMware estate, compares it to performance and cost options across Amazon EC2 with NetApp storage, from which it can suggest a migration plan.

AI partnerships

Also on the AI front, NetApp announced testing will soon start on Nvidia DGX SuperPod and NetApp AFF A90 storage.

Also going GA at Insight 2024 were a NetApp AIPod with Lenovo, comprising Lenovo servers, NetApp storage and the Nvidia OVX platform, plus a new NetApp FlexPod converged infrastructure offer targeted at AI RAG workloads. 

Source is ComputerWeekly.com

Vorig artikelMedtech startup brings Oracle AI to bear on cancer drug research
Volgend artikelPure punts raft of unifying features in FlashBlade file and object