10x Backbone: Scaling Backbone Connectivity to Serve AI Demands - Live from SCC

Meta's backbone engineering team addresses the challenges of scaling network infrastructure to meet increasing demands, particularly from AI workloads. The discussion centers on accelerating data center build-out plans and expanding network capacity. Key strategies include pre-building data center metro components, platform scaling (both up and out), and integrating IP and optical layers using coherent router (CR) technology to reduce power consumption and space allocation. The speakers also explore interconnecting data centers within a 3 to 10-kilometer range, using high fiber counts, and employing DWDM optical systems for longer distances to multiplex connections and minimize disruption. They highlight the scale of AI backbone site pairs, which can be double the size of the entire backbone network, and emphasize optimizing deployment speed and operational practices.