Keeping Your Cloud Footprint in Check
summary
Netflix runs 50–100k instances across three AWS regions with hundreds of microservices that autoscale, using red/black deploys and monthly cross-region failovers; engineering buys heavy three‑year reservations for ~70% discounts and shares capacity via linked accounts and an internal borrowing API. Monitoring and tooling include Atlas: 2 billion time series/min, Asgard, Chaos Monkey/Kong, Vector, hisy for circuit breaking, Kronos for audit and predictive autoscaling. Internal borrowing yields concrete savings like 135,000 instance hours saved (~$200k/month).
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:28About Coburn Watson
- 3:09Open Source
- 4:53Priorities
- 5:12Netflix
- 6:19Innovation
- 7:18Reliability
- 9:32KPI
- 10:37VP Cost
- 19:31Optimization
- 22:26Performance
- 26:37Internal Borrowing
related talks
27mBehind the Scenes: Running Critical Services @ Netflix
47mCloud Security at Scale and What it Means for Your Application
19mKeynote: Making and Keeping Netflix Highly Available
01h 52mDeep Dive into the Cloud Native Open Source with NetflixOSS
50mFast Delivery
32mHow Netflix does failovers in 7 minutes flat
31mHow Netflix Thinks About DevOps
39mCloud Native Architectures