Riot Direct Network - Our infrastructure as Code Journey
summary
Riot Direct codifies device configs as YAML in GitHub, uses Jenkins CI to run Nornir/Napalm-style deploys and Capirca for ACLs, and enforces a master branch = production workflow. The network spans ~32 PoPs, ~300 ISPs and ~2,500 eBGP sessions, targets ~3–5 ms latency and 128 Hz tick servers, and has produced over 2,200 merges since Sept 2019. Deploys run a dry‑run diff posted to the PR then a final push using full config replace, and the system is working and broadly adopted.
AI-written summary · about
chapters (15)
- 0:00Introduction and background
- 1:06Riot Games network challenges
- 4:39Infrastructure as Code concept
- 6:04Key design principles
- 8:11Full config replacement benefits
- 9:15Leveraging existing Riot systems
- 10:18System architecture overview
- 11:43Device properties and templates
- 13:26Exception handling in configs
- 14:28Device versioning approach
- 15:54Change tracking and visibility
- 17:39Workflow with Git and CI/CD
- 23:03Live demo of the workflow
- 27:47Adoption and future plans
- 29:54Key takeaways and conclusion
related talks
34mFixing the Internet for Real-Time Applications (Games)
56mA Practical Look at NetDevOps
37mZero-Regression Network CI/CD for Finance-Grade Reliability
51mOur Journey to a self-healing network
25mNetDevOps: Applying lessons of DevOps to networking
40mNew Network Provisioning System Leveraging Kubernetes and Cloud Native Open Source
19mNetcastle: Network Infrastructure Testing At Scale
23mThe Shift: Radio Network Functions as IT Applications