Learn about scaling AI networking infrastructure to support massive GPU node deployments in this technical presentation from Edgecore Networks' Senior VP of R&D. Explore emerging technologies for scaling out AI networking nodes, including innovations in optics, power delivery, cooling systems, and rack management solutions. Discover how to address various infrastructure challenges and constraints while supporting the growing demands of complex AI applications. Gain insights into practical approaches for managing and scaling networks to accommodate up to 100,000 GPU nodes, with detailed discussion of different scenarios and their corresponding technical solutions.
Overview
Syllabus
The Art of AI Networking Scaling – to support 100K GPU Nodes Presented by Edgecore Networks
Taught by
Open Compute Project
Related Courses
-
Scaling Kubernetes Networking to 1k, 5k, and 100k Nodes
-
Liquid-Cooled Vertical Busbar Design for 200kW AI Computing Racks
-
Infrastructure Software for Hyperscale AI and GPU Clusters
-
AI at the Intersection of Storage and Networking
-
Scaling Systems for Generative AI - Building and Deploying Large Language Models
-
Pushing Infrastructure Boundaries: Memory Requirements in AI Data Centers
