Case Study: boo-box serves 1 billion ads per month and delivers sub-250ms real-time targeting with MySQL

boo-box Servers 1 Billion Advertisements per Month with MySQL and Hadoop

boo-box is one of South America's largest advertising networks, focused on the Brazilian social media market, that connects publishers with advertisers to deliver precisely targeted campaigns. Their challenge was to monetize publisher content and deliver relevant ads in real time at massive scale—handling 2 TB of raw logs per month and billions of rows of data—while meeting strict latency targets (under 250 ms) and keeping costs and operational complexity low.

boo-box built a hybrid stack using MySQL for low-latency event logging and serving (nginx + Lua, supporting 20,000 TPS on a single server) and Hadoop on AWS (S3, EMR, Apache Pig) for large-scale analytics, moving data between MySQL, S3 and BigQuery. The solution processes ~22 billion rows, stores an 8 TB BI database, and serves about 1 billion ads per month with sub-250 ms response times, enabling rapid growth and ongoing improvements (MySQL Applier for Hadoop, MySQL 5.6, and MySQL Cluster under evaluation).

boo-box

Josafá Santos

IT Manager