How We Handle Over 1 Trillion HTTP Requests ...
summary
Criteo runs 11 on‑prem data centers with ~1 million CPU cores, ~30k servers and petabytes of RAM to serve over 5 billion ads per day and absorb 1 trillion HTTP requests (peaking ~19 million queries/sec). Real‑time bidding loops complete under 100 ms; 2,000 ML models deploy daily producing ~120 million predictions/sec, reprocessing ~7 million catalog items/day and running ~300 A/B tests/year. Operational controls include deployment freezes, scaled critical services, thousands of dashboards/alerts, 25 TB logs/day in OpenSearch and on‑call via PagerDuty.
AI-written summary · about
chapters (9)
- 0:00Introduction to ad tech
- 5:55Industry landscape
- 7:57Criteo's platform mechanics
- 10:17Scale and engineering teams
- 12:24Infrastructure strategy
- 16:27Technical stack and architecture
- 20:32Microservices and DevOps
- 23:00AI at scale and Black Friday
- 28:41Sustainability and future goals
related talks
44mHow to Handle Over 1 Trillion HTTP Requests/day: Behind the Scenes of the Ad Tech Industry
48m10 Billion a Day,100 Milliseconds Per: Monitoring Real Time Bidding
01h 02mAchieving High Load in Advertising Technology
11mData science: Next-gen performance analytics
38mCrunching data in real time for Fortune500 clients
26mHow We Prepped Our Infrastructure for a Super-Bowl Sized Traffic
33mAnd now to something ELSE: Real Time Data Processing @ billiger.de
42mBlack Friday Brilliance: Managing a Billion Transactions with Tech, Tactics, and Teamwork