How do you build a cloud service that can compete with Microsoft Azure or Amazon Web Services? For most service providers, it starts with a call to Supermicro, which has quietly emerged as the infrastructure hardware provider of choice for nearly every IaaS/PaaS provider that isn’t a tier-1 CSP.
How do you build a GPU cloud atop your cloud service? That requires another phone call, this time to a provider like VAST Data, which has quickly become the data plane of choice for GPU-cloud providers like Lambda Labs and CoreWeave.
VAST Data and Supermicro have complementary interests and overlapping customers, so last week when the two companies announced a collaboration to deliver a full-stack end-to-end AI solution, I wasn’t surprised. Scratch that: this one’s so natural that I was surprised that it hadn’t happened before.
The Solution
The Supermicro/VAST Data solution leverages NVIDIA-certified systems to provide service providers, technology companies, and large enterprises with a powerful, scalable, and secure platform for AI applications.
The new offering is built around an innovative parallel architecture and unified global namespace ensure optimal GPU utilization, scalability, and smooth data access from edge to cloud, eliminating the usual trade-offs between performance and capacity.
Leveraging Supermicro’s NVIDIA-certified offerings along with VAST Data’s software platform, the collaboration integrates computing and data functionalities onto a Supermicro platform, enhancing deep learning capabilities and suitability for extensive AI deployments.
Key benefits of the offering include:
- Simplified large-scale data deployments: The VAST Data Platform on Supermicro ensures security, scalability, and high performance. It revolutionizes architecture to provide seamless data management and access from edge to cloud, meeting the needs of large enterprises embarking on ambitious AI projects.
- Limitless scalability and high GPU utilization: With systems designed for rapid deployment, the solution can scale to exabyte levels while maintaining 99.999% uptime.
- Linear scalability through parallel architecture: The VAST DASE™ architecture, available on industry-standard servers, offers a parallel system design enabling support for the world’s largest GPU clusters.
- Comprehensive support for data preparation and model training: Enhances feature engineering and supports extensive training and inference workloads.
- Instant data access from edge to cloud: This solution utilizes VAST’s global namespace for immediate data availability and consistent performance across locations.
- Cost optimization: Offers the efficiencies of flash storage at the economics of archive solutions, eliminating the need for data tiering.
- Enhanced security for multi-tenancy: Features customer-managed encryption keys, audit tools, and isolation of data management from the host OS to mitigate risks in shared environments.
Analysis
The VAST Data and Supermicro solution caters to the demanding needs of Cloud Service Providers and large enterprises. It harnesses the strengths of VAST Data’s software platform and Supermicro’s high-performance hardware to offer a scalable, resilient, and efficient system for managing large-scale AI deployments, LLM and other Generative AI workloads.
There’s no deep analysis required for this one. It’s a solution whose distinct parts are well-proven to deliver performant AI at scale in some of the most demanding GPU-cloud environments on the plane. Working together to further optimize what’s available is a move that simply makes sense.
