Leveraging open source dynamic routing to implement a spine/leaf in datacenter
summary
Explains how OpenStack shifts from stretched Layer 2 to a true Layer 3 spine/leaf by advertising controller and VM IPs via FRRouting (FRR) and an OVN BGP agent that creates loopback interfaces and installs routes into the kernel. Uses BFD for sub‑millisecond failure detection and ECMP for immediate multipath takeover; implements proxy ARP/NDP and kernel IP rules to steer traffic to provider bridges. Notes FRR runs in containers on compute/control/network nodes and migration requires downtime or region-by-region cutover.
AI-written summary · about
chapters (10)
- 0:00Introduction to open source routing
- 2:12OpenStack DCN deployment architectures
- 6:27Distributed control plane with BGP
- 10:17Layer 3 data plane processing
- 16:22Free Range Routing (FRR) overview
- 19:08BGP in spine/leaf architectures
- 21:20BFD for fast failure detection
- 23:30ECMP for link redundancy
- 27:07References and resources
- 28:35Q&A: Migration challenges
related talks
22mLayer 3 networking with BGP in hyperscale DCs
30mHow we integrated L3 services into one of the biggest OpenStack Public Clouds - OVHcloud case study
19mCrossing the border - VM goes on a BGP adventure
34mHyperRouter: Lessons Learnt from Building an L4 Load...
39mUsing BGP at OpenStack to interconnect workloads across clouds
19mStarfish: A Topology-Routing Co-Design for Small-Scale Data Centers
41mImpact of Network Automation
29mLiving on the Edge with DCN Networking