One of their specific design objectives was to ensure Netflix services kept running, even if an entire AWS availability zone went down, such as happened with US-EAST. To do this required that their system be loosely-coupled, with each component having aggressive timeouts to ensure that failing components didn’t bring the entire system down.Instead, each feature and component was designed to gracefully degrade. For example, during traffic surges that created CPU-usage spikes, instead of showing a list of movies personalized to the user, they would show static content, such as cached or
...more
