Equinix Speeds Distributed AI Infrastructure Applications with NVIDIA

As the amount of data generated at the edge increases, it is prudent to process it at the source – instead of moving massive datasets to a remote centralized location for cost, latency and privacy reasons. In addition, most AI applications need to access multiple external sources of data that are spread across public clouds, private data centers, data brokers and from internet of things (IoT) devices at the edge. So, it makes sense to host the artificial intelligence (AI) stack at an interconnection hub for greater performance and secure access to these numerous data sources.

AI and data gravity go hand-in-hand. As data accumulates, additional AI services and applications will be attracted to this data, making it vitally important to choose an optimal location for compute infrastructure. Moreover, in many use cases, for compliance reasons, data must be processed in the country of origin, requiring AI stacks to be located in multiple countries. These trends precipitate the need to perform AI model training and inference in a distributed manner. It is for these reasons that Equinix and NVIDIA are offering customers the NVIDIA LaunchPad “instant AI infrastructure” solution on our globally distributed Platform Equinix®.

How Equinix and NVIDIA accelerate distributed AI infrastructure applications

The NVIDIA LaunchPad program provides access to optimized NVIDIA software running on accelerated infrastructure at nine Equinix locations worldwide. In LaunchPad, qualified customers can prototype and test data science and AI workloads using the NVIDIA Base Command cloud service for AI development, the NVIDIA Fleet Command cloud service for AI deployment or the NVIDIA AI Enterprise software suite for end-to-end AI workflows on VMware vSphere. LaunchPad features NVIDIA-Certified Systems running on Equinix Metal™ (which is also offered as a service in multiple Equinix locations), and NVIDIA DGX systems, providing accelerated infrastructure for AI training, AI inference and data science.

Equinix Fabric™ software-defined interconnection delivers high-speed and secure connectivity between these distributed training and inference locations, and the various distributed data sources. In addition to the AI compute, network and storage infrastructure, NVIDIA LaunchPad with Equinix Fabric provides the necessary software-based orchestration services to move data and AI models between the distributed sites in a seamless manner. This total solution is further detailed in our article titled Equinix and NVIDIA AI LaunchPad Accelerate AI from Hybrid Cloud to Edge.  Customers can use NVIDIA LaunchPad to try out these solutions and subsequently procure and deploy NVIDIA hardware in their own private cage at Equinix.

NVIDIA AI Enterprise, Base Command and Fleet Command are ideal for end-to-end AI development to deployment:

• Frameworks and tools that are optimized, certified and supported by NVIDIA to enable the rapid deployment, management and scaling of AI applications on mainstream NVIDIA-Certified Systems and VMware vSphere environments, deployed at Equinix data centers or on-premises.
• Base Command – A cloud-based software for businesses and data scientists to execute AI development workloads on accelerated AI infrastructure based on the NVIDIA DGX SuperPOD architecture, hosted at Equinix. It manages the end-to-end lifecycle of AI development, includes workload management and resource sharing, and has an intuitive graphical user interface and a command line interface, as well as integrated monitoring and reporting dashboards.
• Fleet Command – A cloud service that securely deploys, manages and scales AI applications across distributed edge infrastructure. It runs on mainstream, accelerated NVIDIA-Certified Servers, deployed at Equinix and enabled by Equinix Metal.

NVIDIA LaunchPAD users can leverage the Equinix Fabric network to move data and models between distributed AI training and inference deployments. NVIDIA LaunchPad allows companies to test the combined AI value proposition of NVIDIA and Equinix, and subsequently procure and deploy this solution on Platform Equinix. Businesses get the following value proposition from this solution:

• A rich, interconnected ecosystem: Users can leverage data from an ecosystem of 10,000 companies, which house their infrastructure at Equinix while building their AI models.
• Proximity to edge devices and clouds: Equinix data centers are strategically located in metros where they are within 10ms round-trip time from most end devices and within 1 to 2ms from most cloud providers. This allows customers to perform near real-time model inference for many use cases, and get high-speed access to data located in the public clouds.
• Access to virtual network devices: Network Edge from Equinix provides virtual network services (virtual firewall, router, etc.) from leading providers, which can be deployed quickly as part of distributed AI architecture. Organizations can select, deploy and connect virtual network functions devices at the edge in minutes, with no additional hardware requirements.
• Interconnected, automated bare metal as a service: Equinix Metal enables organizations to deploy physical infrastructure at software speed. Using an OPEX model, businesses can deploy powerful, interconnected bare metal infrastructure on Platform Equinix in minutes, across global locations.
• Global interconnected presence: Equinix data centers are present in 65 metros in 27 countries and are interconnected via a high-speed and secure network fabric. Organizations can do business with a single data center and network fabric vendor as they deploy their AI stacks in multiple countries to maintain regulatory compliance.
• Interconnected digital infrastructure: Equinix Fabric software-defined interconnection provides high-speed and secure connectivity between distributed training and inference locations, and the various distributed data sources.

For more information, read the Equinix and NVIDIA white paper titled “Artificial Intelligence: From the Public Cloud to the Device Edge.”

You can also attend the Equinix session, “AI Development to Development on Platform Equinix,” at NVIDIA GTC 2021, running Nov. 8-11.