NVIDIA’s strong channel focus ensures that this joint offering with DDN will be completely channel-centric, although the number of partners certified on the NVIDIA DGX-1 and with DDN is fairly limited at this point.
Today, DataDirect Networks [DDN] is announcing a new partnership with NVIDIA. It will see an integration between DDN’s A³I platform and the NVIDIA DGX-1 AI supercomputer, which will produce a validated and pre-configured offering that is optimized for AI and deep learning workloads. It will go to market entirely through a fairly select channel, that is certified on both DDN and the DGX-1, the latter of which has a very rigorous certification process,
“NVIDIA is very aggressive in these types of partnerships, and partners with a lot of strategic vendors,” said Kurt Kuckein, DDN’s senior director of marketing. “Where we think we have differentiation is our ability to approach the solution from a different angle. We use a parallel file system architecture rather than NFS, which allows us to deliver greater performance as the solution scales, and to extract more capabilities from the DGX-1, with full GPU resource utilization. Other vendors emphasize the way their DGX-1 integration delivers really fast storage. However, our goal isn’t just to provide fast storage, but to enable users to be able to do more work, and execute better analysis. The way our file system interacts with the NVIDIA GPUs is a net enabler of these tasks.”
This is the first time that DDN has partnered with NVIDIA.
“We had met them in the field on multiple occasions,” Kuckein said. “That’s to be expected, because we are heavily into HPC and that scales through GPUs. There was mutual interest in partnering based on the fact that we were starting to drive AI as a message, and they have been a thought leader here from the compute side. Our focus around AI and deep learning has been mainly on the HPC side though. With this partnership, we will be able to expand our opportunity on the commercial side. This partnership will provide opportunities for a ready-to-use rack level solution designed for a particular workload using a jointly-designed reference architecture.”
DDN’s part of the equation, the A³I, is not a flagship offering, but a purpose-built platform designed to be customized to particular AI hardware-based purposes.
“The A³I is a family of products specifically developed for AI and deep learning workflows,” Kuckein said. “We take existing hardware products – in this case the NVIDIA DGX-1 environment – and tune specifically for them.”
DDN sees a couple of sweet spots for this joint offering.
“The reason this product now exists is that both NVIDIA and ourselves were finding customers who had kicked off a new AI project, and who had hired a couple of data scientists who want a specialized infrastructure,” Kuckein said. “With this, we have a turnkey solution, where the customer doesn’t need to know anything about the GPU environment, and it is something that they are able to deploy easily at the rack level. We are also seeing another use case, among folks who have implemented an AI architecture before. However, they are finding that they are spending too much time and money working on keeping storage and compute running optimally instead of on the problem that they designed the architecture to solve.” With our background in high performance storage, we understand how to start small and scale, so can help these customers with an easy button that is also scalable over time.”
The go-to-market motion for this offering is new for DDN. While they do have a channel, this offering will be sold through a more channel-centric model than DDN has typically used in the past, and that’s a reflection of NVIDIA’s channel-centric go-to-market.
“We have a large number of reseller partners, although we focus on a smaller niche of them who specialize in the same markets we do – the federal space, the university research community, financial services, and oil and gas – where we can bring value to each other,” Kuckein said. “However, we have had a fairly direct sales engagement historically, which uses resellers for fulfilment. This offering is different because it has a channel-first focus, and is specifically intended to enable partners more. It lets them sell the AI solution, and provide the Level 1 and Level 2 support, and not have to rely on DDN to be hands-on with the customer. That’s NVIDIAs model – it’s all channel enablement. So to go to market with NVIDIA, we had to increase the robustness of our capabilities on the channel side. We have been working over the past year to do a better job of enabling partners. We upgraded our portal in July, and are fully populated with this solution there.”
Still, Kuckein said that even though this is a turnkey solution, it’s not something that all DDN partners will be able to sell. Only a few partners have been identified as able to sell it at launch, including Meadowgate Technologies, Microway and Groupware Technology in the U.S., and GDEP Solutions, XENON and E4 Computer Engineering elsewhere.
“That’s because in addition to being a DDN partner, they have to have received full training, be capable of doing the Level 1 and Level 2 support and have been certified by NVIDIA on the DGX-1,” Kuckein said. “That’s the extra step for most, the DGX-1 certification. Many of our partners can resell discrete GPU solutions from NVIDIA, but the DGX-1 has a much more extensive certification program.”
The DDN A³I with NVIDIA DGX-1 solution is available now.