Netflix’s 57M members watch over 2 billion hours of content per month and their streaming accounts for 1/3rd of Internet traffic in some parts of the world. Each time a member starts to watch a movie or TV episode, a “view” is created in our data systems and a collection of events describing that view is gathered. The system Netflix built to manage and process this critical data devolved over time into a monolith, but needed to evolve into microservices to be able to scale to the next order of magnitude.

In this talk, we’ll describe how the team re-designed and evolved a critical system of this scale while still “keeping the lights on” (processing a billion events per day). Architectural patterns discussed will include service decomposition, stateless application tiers, and polyglot persistence. Migration strategies discussed will include shadow writes, KPI comparisons, feature dials, and fast rollbacks. Attendees will learn patterns and techniques that can be applied to architectural evolutions that they undertake.