How Netflix enables
summary
Netflix runs a self‑service CockroachDB-as-a-Service with an automated control plane (Spinner/anti-gravity) and ~380+ clusters, ~160 production and 60 multi-region clusters, including a 48‑node, four‑region gaming cluster. Multi‑region is sold for region survivability not load shedding — extra regions increase cross‑region traffic and can hurt performance. Default topology uses zone configs with three replicas per AZ (three regions → nine replicas); practical tactics include leaseholder pinning, follower reads, geo‑partitioning and gateway nodes when adding regions.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 0:33Behind Netflix's data platform team
- 3:24What does Netflix's fleet look like?
- 5:13Balancing during traffic evacuation: Advantages of multi-region databases
- 7:28Region survivability: Advantages of multi-region databases
- 8:23Performance: Advantages of multi-region databses
- 8:56Performance considerations for a multi-region setup
- 12:42How to design around multi-region requirements
- 15:19Topology Configurations (Abstraction vs Zone configurations)
- 17:00Default multi-region topology configurations
- 18:42How to expand to new regions
- 19:48Topology configuration examples
- 21:36What's next? (Abstraction level customer coaching and more)
related talks
44mTechniques Netflix Uses to Weather Significant Demand Shifts
39mAvailability, Latency, and Cost: Withstanding Regional Outages
46mCassandra Serving Netflix @ Scale
33mWhat We’ve Learned Building a Multi-Region DBaaS on Kubernetes
42mHow Netflix Built a Real-Time Distributed Graph for Internet Scale
14mHow CockroachDB can enable Multi-Region Deployments and Data Locality 
34mHow Netflix Delivers Key-Value and Time-Series Storage at Any Scale
26mBuilding Financial Systems on Eventually Consistent DBs