Nutanix beefs up AHV hypervisor, doubles down on AI

Nutanix takes aim at VMware with enhancements to its AHV hypervisor along with tighter integrations with Nvidia and Hugging Face to simplify AI deployments

Nutanix is making it easier for organisations to move away from VMware and deploy artificial intelligence (AI) workloads through a slew of enhancements unveiled at its Next 2024 conference in Barcelona last week.

It announced Nutanix AHV deployment options with partners such as Cisco to certify Cisco UCS blade servers to run AHV, giving customers the flexibility to use the AHV hypervisor on compute-only or storage-only nodes. Nutanix is working with other partners to broaden the range of new and existing server configurations certified for AHV.

Furthermore, Nutanix is now supporting the repurposing of many of the most popular vSAN ReadyNode validated server configurations to run Nutanix Cloud Infrastructure with AHV.

Together, the changes can help customers lower the total cost of ownership (TCO) when modernising their infrastructure, according to the company.

“We are excited to work with our partners to expand the reach of Nutanix AHV to compute-only servers beyond traditional hyperconverged servers, further accelerating its adoption by enterprise customers to simplify operations and increase cyber resilience,” said Thomas Cornely, Nutanix’s senior vice-president of product management.

“This year celebrates the 10th anniversary of the launch of Nutanix AHV hypervisor with a continued focus on innovation to deliver the best enterprise virtualisation and containerisation platform,” he added.

The resilience of AHV-based systems is also being improved by new multi-party approval controls for changes to Secure Snapshot to protect against ransomware and threat actors, and forthcoming support for multi-site disaster recovery to minimise workload downtime during outages.

Easing generative AI deployments

The second iteration of Nutanix GPT-in-a-Box, which was first launched last year, includes integrations with Nvidia Inference Microservices (NIM) and easy access to Hugging Face’s large language model (LLM) library. Other changes include a new graphical user interface, end-user access key management, integration with Nutanix Files and Objects, and the ability to use Nvidia’s Tensor Core graphics processing units (GPUs).

“We saw a great response to our original launch of Nutanix GPT-in-a-Box, validating the needs of enterprise customers for on-premises software solutions that simplify the deployment and management of AI models and inference endpoints,” said Cornely.

“Enterprise is the new frontier for GenAI [generative AI] and we’re excited to work with our fast growing ecosystem of partners to make it as simple as possible to run GenAI applications on-premise at scale while maintaining control on privacy and cost,” he added.

Manosiz Bhattacharyya, Nutanix’s chief technology officer, said the integration with NIM will simplify AI model deployment across the enterprise and at the edge, adding: “There’s no point creating models every year – the challenge is to use them.”

Clem Delangue, founder and CEO of Hugging Face, said: “Our partnership with Nutanix will help extend Hugging Face to more enterprises looking for private and secure control of their LLMs. Our mission is to enable organisations of all sizes to build their own AI with open source. Our work with Nutanix will make it easy for enterprises to build AI applications on-premises, leveraging text generation inference.”

To better support AI, machine learning and other applications, Nutanix Unified Storage now has the option of a new dense low-cost all-NVMe platform using Micron’s 30TB flash modules, with up to 10 GB per second sequential read throughput from a single node. Nutanix will also add support for Nvidia GPUDirect Storage for additional performance improvements.

AI-optimised GPUs, including the Nvidia L40S, H100, L40, and L4 GPUs, are now supported on Nutanix GPT-in-a-Box, as well as density-optimised GPU systems from Dell, HPE, and Lenovo to help reduce TCO by using fewer systems to meet their workload demands.

“Across every industry, enterprises are working to efficiently integrate AI into cloud and data platforms that power their operations,” said Manuvir Das, Nvidia’s vice-president of enterprise computing. “The integration of NIM into Nutanix GPT-in-a-Box gives enterprises an AI-ready solution for rapidly deploying optimised models in production.”

The new version of GPT in-a-Box is expected to be available in the second half of 2024.

Read more about AI in APAC

Read more on Platform-as-a-Service (PaaS)

Data Center
Data Management