Building scalable AI cloud interconnects with SRv6
summary
Nebius replaces stitched VXLAN+MPLS stacks with an SRv6-first design that provides end-to-end Layer‑3 VPNs and traffic engineering across data center and WAN domains, handling customer demands in the hundreds to multiple terabits/sec. Implementation uses the VPP data plane for SRv6 encap and proves SRv6 interop across vendors, while SRv6 micro SID compression reduces packet overhead for software routers. Service chaining starts with SRv6‑unaware firewalls via proxy nodes and promises true scale once firewall code and a traffic‑engineering controller are added.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:52Nebius network architecture
- 4:18Cloud interconnect requirements
- 6:07Why SRv6 fits
- 8:16SRv6 micro segment overlay
- 10:46SRv6 firewall chaining
- 12:11Implementation experience
- 13:56Future SRv6 plans
related talks
27mLeveraging Cilium and SRv6 for Telco Networking
27mSimplifying and Making the Network Programmable with Kubernetes and SRv6
39mSRv6LB: Leveraging IPv6, Segment Routing, and VPP
22mLayer 3 networking with BGP in hyperscale DCs
29mLiving on the Edge with DCN Networking
30mHow we integrated L3 services into one of the biggest OpenStack Public Clouds - OVHcloud case study
19mBluebird: High-performance SDN for Bare-metal Cloud Services
23mAIN: A Blueprint for an All-IP Data Center Network