Главная
Study mode:
on
1
Intro
2
What is an LPO (Linear Pluggable Optic)?
3
Comparison of Optical Interconnect Technologies
4
Enhanced Interconnect for Al Fabric - Summary
5
Tomahawk Al Custer Architecture - Endpoint Scheduling
6
Enhancing Ethernet Bandwidth
7
Tomahawk 5: 800GE and Al Workload Acceleration
8
Tomahawk 5 - Benefits of Enhanced Bandwidth
9
Tomahawk 5-Broadcom Cognitive Routing
10
Tomahawk 5 Cognitive Routing
11
Tomahawk 5 Global Load Balancing (GLB)
12
Tomahawk 5 Reactive Path Rebalancing
13
Tomahawk 5 Drop Congestion Notification (trimming)
14
Example 64K-GPU Cluster with Tomahawk 5
Description:
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only! Grab it Watch a technical presentation from Networking Field Day 32 exploring Broadcom's revolutionary Tomahawk AI interconnect solution, which delivers optimized networking capabilities for hyperscale environments. Dive deep into the architecture of Tomahawk Fabric's high-bandwidth, high-radix design that enables flatter network topologies and enhanced performance through advanced features like load balancing, traffic spraying, and comprehensive telemetry. Learn about the implementation of fast link failover mechanisms, customizable congestion management, and the groundbreaking Tomahawk 5 with native 800 GigE port support. Explore detailed technical aspects including Linear Pluggable Optics (LPO), optical interconnect technologies, AI cluster architecture, cognitive routing, global load balancing, reactive path rebalancing, and drop congestion notification. Understand how these innovations come together in a practical 64K-GPU cluster implementation, demonstrating Broadcom's commitment to meeting the demanding requirements of modern AI architectures. Read more

Tomahawk 5 AI Interconnect and Network Fabric Architecture

Tech Field Day
Add to list