Case Study: Netflix Microservices Architecture

In August 2008, a database corruption took Netflix down for three days. At the time the company still mailed DVDs, and the failure meant it could not ship discs to members for most of a week. The cause was a single relational database in a single data center, the kind of vertically-scaled box where one bad day takes the whole business with it.

That incident convinced Netflix that the path forward was not a larger database in a stronger data center but the opposite: many smaller pieces, spread across cheap commodity infrastructure, where no single failure could take the system down. The migration that followed, from a monolithic data center to a thousand-plus services on AWS, started in 2009 and took roughly seven years. Along the way Netflix invented or popularized much of the vocabulary of microservices: circuit breakers in practice, chaos engineering, client-side load balancing, and the principle that you assume failure rather than try to prevent it.

That is why Netflix is the canonical reference in microservices interviews. This chapter gives you the story behind the patterns, so you can cite it with judgment rather than as a name to drop.