Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore lightweight traffic engineering solutions for AI/ML data center fabrics in this 50-minute conference talk from NANOG. Learn how modern data centers using CLOS IP fabrics with EBGP face unique challenges from AI/ML workloads that demand low-latency, lossless transport while generating high-throughput, low-entropy flows that create congestion hotspots. Discover how BGP Deterministic Path Forwarding (DPF) and BGP Global Load Balancing (GLB) provide effective solutions to these challenges, with DPF enabling deterministic IP routing for predictable flow distribution and GLB offering sub-millisecond congestion mitigation across the fabric. Understand how these mechanisms can work independently or together to enhance AI/ML traffic performance and reliability in large-scale deployments. Gain insights from Kevin Wang, a Distinguished Engineer at Juniper Networks with 16 years of experience in BGP and MPLS areas and major innovations including BGP rib-sharding, FIB Compression, and AI/ML fabric Global Load Balancing, and Michal Styszynski, a Product Management team member at Juniper Networks with over 10 years of experience in data center and storage networking projects for major telcos and enterprises.
