NVIDIA unveils first cloud-native, multi-tenant supercomputer

At GTC21, NVIDIA unveiled the next-generation NVIDIA DGX SuperPOD, the world’s first cloud-native, multi-tenant AI supercomputer. It features BlueField-2 data processing units that offload, accelerate and isolate users’ data to provide users with secure connections to their AI infrastructure.

DGX SuperPODs are AI supercomputers come with 20 or more NVIDIA DGX A100 systems and NVIDIA InfiniBand HDR networking.

They are designed to meet growing security and scalability requirements as AI broadens in adoption. Increasingly, enterprise IT departments need to support the work of multiple teams at different locations, while academic and research institutions often grant outside organisations access to their computing resources.

BlueField-2 DPUs offload, accelerate and isolate users and their data, allowing organisations to safely provide private cloud access to their DGX SuperPOD infrastructure with security that spans user traffic, firewalls and multi-tenant access to storage.

Among the early adopters in Asia are South Korea’s Naver, Japan’s Sony and Vietnam’s VinAI.

“The new DGX SuperPOD, which combines multiple DGX systems, provides a turnkey AI data centre that can be securely shared across entire teams of researchers and developers,” said Charlie Boyle, Vice President and General Manager of DGX systems at NVIDIA. “

NVIDIA has also announced NVIDIA Base Command, which enables multiple users and IT teams to securely access, share and operate their DGX SuperPOD infrastructure. Base Command coordinates AI training and operations on DGX SuperPOD infrastructure to enable the work of teams of data scientists and developers located around the globe. This allows teams of AI developers and data scientists to seamlessly provision and schedule workloads on DGX infrastructure from prototyping to production.

To support the iterative process of building and refining AI models, Base Command provides built-in telemetry for users to validate deep learning techniques, workload settings and resource allocations to constantly improve results.

NVIDIA DGX SuperPODs and NVIDIA Base Command will be available in Q2.