Within Teads, the Buying Engine is the backend responsible for delivering thousands of Ad campaigns at all times through hundreds of thousands of opportunities per second, without ever exceeding our users’ budgets. With traffic growth, the service was having difficulty scaling horizontally, forcing us to move away from traditional architecture patterns.
In this talk, I propose to demystify communication in push mode with gRPC, which allowed us to divide by 4 the load on our Cassandra cluster, and by 1000 the data flow for budget control, while maintaining the same level of service.
