Both partnerships expand AI capabilities by combining VAST’s software with NVIDIA and Super Micro’s hardware, with more such partnerships on the way.

VAST Data, a storage provider which is increasingly defining itself as a data platform company, has launched two important new partnerships. The first uses VAST Data’s parallel file software on Nvidia’s BlueField-3 data processing units (DPUs) hardware, giving NVIDIA their first software to run on Bluefield. It is now being deployed by CoreWeave, a GPU cloud service provider and a mutual customer. The second VAST partnership is with Supermicro on a full-stack AI offering, from storage to compute. It is based on a new reference architecture that provides software and hardware combinations for AI workloads that use VAST software and SuperMicro hardware. SuperMicro is the first OEM to receive this capability, which is also certified by NVIDIA, and VAST says it is just the first of the OEMs that will see this happen.

“Last year was a banner year in the AI market, and we saw great tradition riding this wave through AI,” said John Mao, VP of Global Business Development at VAST Data. “We are now seeing a tremendous amount of large customers using VAST, and we have added gesture recognition and other AI capabilities around speech to text. Pixar has used us to build their last 3 feature films, which had new types of AI algorithms. We have been working with them over 3 years now. Zoom is a customer. We also now have the largest AI, provider in CoreWeave, which has been a tremendous success overseas. Every major country is putting money into sovereign clouds.”

Mao said that on the heels of that, VAST has been working feverishly on the product side. While the initial focus in putting their software on hardware began with support in 2022 for the Ceres hardware platform, their plan now is to expand that number of hardware platforms that it supports.

“In February, we announced our software partnership with Run:AI software partnership redefining AI operations at scale by offering a full-stack solution encompassing compute, storage, and data management,” Mao said. “We started this wave leading up to the NVIDIA GTC 2024 conference, where we will also be emphasizing these announcements with NVIDIA and Super Micro.”

Mao noted that while NVIDIA has had Bluefield around for a while, they now face competition for its DPU technology.

“Amazon has a proprietary DPU with new capabilities around security, QOS and  performance,” he said. “It’s hard to compete with. We have built an optimized version of our software to be deployed on top of Bluefield. NVIDIA will use this to fast track things, but they haven’t had any software to run on Bluefield. Now with us they have a full AI infrastructure which combines both hardware and software, and we can provide an end-to-end BlueField solution.”

The idea behind this partnership is to industrialize AI computing. VAST’s Disaggregated, Shared Everything [DASE] architecture leverages the processing power of NVIDIA BlueField-3 to require less independent compute and networking resources, reducing the power usage and data center footprint for VAST infrastructure by 70%.

By providing each GPU server with a dedicated and parallel storage and database container, this new AI Factory architecture improves QOS [Quality of Service] and eliminates contention for data services infrastructure. The parallelism lets each NVIDIA BlueField-3 can read and write into shared namespaces of the VAST Data Platform without coordinating IO across containers., allowing multi-tenant service providers to meet the contractual Service Level Objectives of their clients while also maximizing the utilization of all GPU computing assets.

The new architecture also enhances Zero-Trust Security, to ensure that data and data management remain protected and isolated from host operating systems. Compared to AI computers that use parallel file system, VAST is able to eliminate many attack vectors in a multi-tenant environment by hosting industry-standard network attached services, object services, and database services through standard client protocols like NFS, SMB, S3 and Apache Arrow from NVIDIA BlueField-3 DPUs that do not expose the underlying Data Platform system topology.

In addition, VAST systems, powered by the NVIDIA DOCA software framework, now provides block storage services natively to host operating systems.

The Super Micro partnership will see VAST and Super Micro jointly equip service providers, hyperscale tech companies and large, data-centric enterprises with a highly performant and resilient AI solution. Organizations can scale VAST and Super Micro clusters to exabyte levels while enjoying over 99.9999 percent uptime, removing the tradeoff between scale and uptime.

“Super Micro is riding that wave as well,” Mao said. “We reached out because of a joint customer about nine months ago. We have built a solution with our software on their hardware for larger customers. There is great synergy there. We have been working on a joint reference architecture, and Super Micro will be the first OEM to provide this.”

Super Micro is distinguished in the industry by their monstrous array of SKUs, and they work with many vendors

“We don’t know what they are doing with other vendors, but they are leaning in very heavily with us,” Mao said. “They do embrace SKUs, but VAST will be in their upper echelon of top tier partners that they move with going forward.”

Mao emphasized that while everyone and their brother in the IT space is climbing aboard the Generative AI bandwagon, VAST has the credentials and experience to back it up.

“ Everyone is doing that now to make them feel more relevant,” he said. “If you don’t do it, you become not relevant. But talk is cheap. It’s easy to put marketing up. We can put up our install base, with customers like CoreWeave, We are their data infrastructure. They have tens of thousands of servers in their data centres running on top of VAST.

“Unlike other companies, we have been doing this for six years,” Mao added. “More than 60% of our workloads in the last four years have been AI and HPC workloads, We aren’t just talking about it.”

Mao said that this is part of a process that is transforming VAST Data as a company.

“We are known as a storage company today, but with the VAST DataBase, the data engine coming will allow a framework for full inference and processing. By this time next year, the world will see we are a data platform company, not a storage company.”