Tools and Team for 20 Billion PageView/Month
summary
Global adtech scales to about 20–24 billion pageviews/month with 1.2B monthly uniques, ~1,600 VMs, 10M requests/min and ~5 PB/month of HTTP traffic. Infrastructure runs primarily on AWS with selective GCP BigQuery use; Terraform codifies 33 regions and reproducible stacks and the Terraform repo with 57 contributors lets devs create infra via pull requests. CI is Jenkins-as-code (700–800 jobs), Docker serves as immutable packaging with one container per host, Chef provisions stateful pets (Cassandra, Kafka, MySQL, Redis) and a six-week project horizontally scales Graphite. Teams are feature-owned for run, on-call uses paired rotation via PagerDuty and Slack.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:57Teads Infrastructure
- 3:05Cloud Native
- 6:17Cloud Monogamy
- 9:14SAAS for the win
- 10:44Monitoring observability
- 14:54Automate all the things
- 17:23Docker all the things
- 19:41Cattle vs Pets
- 21:48Industrialize
- 26:03Feature Teans
- 28:32Infra Team
- 36:05Takeaways
- 38:14We're Hiring
related talks
27mScalable by Design: How to Build an Elastic Cloud Infrastructure...
47mScaling: from 0 to 25 million users
51mAgency Workflows
46mManaging a Billion Transactions with Tech, Tactics, and Teamwork
35mScaling Innovation: ADB's Cloud Journey
54mGetting to 300 million requests a month with a one man AI/Dev-Ops team
23mIt Takes a Nation of Millions to Hold Us Back
14mHow Global Discovery tribe approaches Infrastructure Changes at scale