Matthew Landry & Murali Gandluru, Cisco | Networking for AI Summit
summary
Claims that backend training traffic now outpaces front end workloads and that multi–data center GPU clusters force new fabric designs are backed by concrete examples: a customer running 16,000 GPUs, emphasis on prioritizing GPU-to-GPU flows, and a cited 93% who view the network as critical. Recommended tools and innovations include AI Canvas for end-to-end visibility, HyperShield for distributed enforcement, liquid-cooled switches, Wi‑Fi 7 for low‑latency inferencing, and agentic operations plus unified observability to enforce security and predictable latency across LAN, WAN, campus and data center.
AI-written summary · about
chapters (9)
- 0:00Intro
- 2:08Enhancing Enterprise Network Security and Intelligence
- 6:38Changes in Network Design with AI
- 10:21Data Center Perspectives on AI Networking
- 14:08The Importance of Network Resilience
- 18:13Reliability and Performance Priorities in AI Networking
- 22:00Innovative AI Use Cases Across Industries
- 26:00Similarities and Differences in AI Networking Needs
- 28:44Strategizing Tomorrow: Crafting a Future-Ready AI Networking Vision
