Highlights:

  • The Nvidia BlueField networking platform establishes a software-defined accelerated computing infrastructure specifically designed for AI.
  • By integrating Vast’s services into the BlueField DPUs, businesses can significantly expedite the transfer of data between AI processing and storage.

A startup specializing in artificial intelligence data platforms, Vast Data Inc., has partnered with Nvidia Corp. and Supermicro Computer. It has unveiled a novel AI cloud architecture in collaboration with Nvidia Corp. This innovative approach utilizes Nvidia’s BlueField-3 data processing unit technology to host Vast’s operating system, transforming supercomputers into powerful AI data engines.

Additionally, the collaboration with Supermicro Computer Inc., a leading global provider of computing hardware solutions for AI, cloud, and storage, is significant. This partnership aims to streamline the deployment of ull-stack, end-to-end AI solutions in large volumes. Vast Data and Supermicro are enabling service providers and hyperscale technology firms to construct data-centric AI-powered solutions using systems integrated with Nvidia DPUs.

The Nvidia BlueField networking platform offers a software-defined accelerated computing infrastructure tailored for AI. This is achieved through the amalgamation of computational capabilities and integrated hardware accelerators designed to fortify networking environments. By incorporating Vast’s operating system into each DPU within a graphical processing unit server, Vast can efficiently parallelize its services, enabling the embedding of storage and databases on an exceptionally large scale.

The Vast’s Data Platform is a software infrastructure framework that gathers, organizes, and maintains unstructured data for artificial intelligence (AI) processes. These processes require the synthesis of large volumes of data for training, optimizing, and deploying in order to handle user queries. According to John Mao, Vice President of business development and technology alliances at Vast Data, companies can significantly speed up the flow of data between AI processing and storage by integrating Vast’s services onto the BlueField DPUs, as he explained to a leading media house in an interview. There are numerous advantages to this for training, deploying, and growing AI applications in data centers, which call for sizable GPU clusters for AI training and deployment activities.

“The benefit for most organizations is that it radically reduces the amount of equipment required. So, there’s obviously context savings. But I think the more interesting thing is that these big GPU clusters if you’ve been following, they’re very power-hungry to take up a lot of power in data centers. It’s not it doesn’t sound like a lot where we did a, we did an analysis that looks at basically about 5% of total datacenter power savings,” Mao said.

Mao noted that while that might not seem like much, a lot of these data centers operate on multimegawatt scales. This implies that over the course of months or a year, a 5% savings can result in significant power savings.

Another significant advantage, he noted, is that by deploying the software on BlueField DPUs, it becomes simpler for AI data center developers to scale out rapidly without requiring any reengineering. “If you start with something relatively small, a few 100 GPUs, as you add, you know, 1,000 GPUs and then 10,000 GPUs, this new architecture allows those clusters to be able to scale performance of those new storage services completely linearly with no thinking and no rearchitecting,” Mao said.

Vast already collaborates with CoreWeave Inc., a cloud computing infrastructure provider powered by GPUs that specializes in AI deployment and training. Vast’s operating software is currently being used in production on Nvidia BlueField DPUs by CoreWeave Inc.

Additionally, the business announced a collaboration with Supermicro to offer full-stack solutions for the deployment of large-scale AI infrastructure. Mao revealed for the first time a Vast Data Platform offering that will be hosted on industry-standard servers. For the company, this has been a paradigm shift as it was operating on more specialized hardware. Originally designed to house the Vast software, customers can now manage their own supply chains on much larger scales using industry-standard hardware.

For the past year, Mao stated that customers have been doing business with Vast in the petabyte and exabyte deployment sizes, and in certain instances, the multiple exabyte scales. Operating at such scales, customers are less concerned about the failure of an individual server. They are far more worried about a whole rack or a collection of servers failing at once.

“Their concept of what they’re optimizing for is different. They also don’t have an appetite — those customers running at that scale — they need to be able to optimize their own supply chains from a hardware perspective,” said Mao.

Companies will find this particularly beneficial, as the Supermicro offering includes the BlueField DPUs integrated onboard rather than being sold separately by Nvidia. These DPUs are an integral part of the technology stack shipped with clustered GPU servers. Hence, Mao mentioned that hyperscale data center providers collaborating with Vast can also leverage this integrated solution.