Three days without DVDs started Netflix's seven-year rebuild

transcript

August 2008. Netflix can't ship DVDs. For three days. A corrupted database. The fix took seven years.

Netflix looked past the database. The enemy: single points of failure. One big machine you keep upgrading. It breaks, everything stops. The goal: many ordinary machines sharing the work, in the cloud. Lose one? It doesn't matter.

They picked AWS. Then came the choice behind the seven years. The easy route: forklift every system into AWS, unchanged. But as Netflix put it, the problems move along with it. So they rebuilt virtually all of their technology. Cloud-native. And fundamentally changed the way they operate.

Seven years of moving. And the business never stopped growing. Eight times as many streaming members. Viewing grew about a thousand times in eight years.

Early January 2016. The last bits of the data center shut down. That same month, Netflix launched in over 130 new countries.

Cost wasn't the main reason. Yet cost per streaming start became a fraction. The old design was the problem. Moving it would move the problem too.

Full breakdown in the lecture.