Hacker News Reader

Comment by hintymad | Hacker News Reader

What I don’t get about Amazon is that AWS customers have way less oncall load, particularly Netflix, whose engineers could afford oncall 24x7 for weeks or months with perfectly normal work-life balance. Why can’t Amazon, the pioneer of cloud computing, achieve the same level of effectiveness?

View on HN

Maybe I'm misunderstanding you but AWS customer's eningeers have less oncall load because the AWS engineers are doing so much of the difficult and failure-prone work.

That's largely what you're buying when paying for cloud services.

What I meant was that Netflix built their platforms and services on top of AWS services, just like AWS teams. Yet AWS teams have brutal oncalls, while Netflix teams enjoy great work-life balance.

Company culture and values flow from the top

But that’s because the AWS teams are presumably shouldering the burden, then.

I’m sure you’re right for some services, especially infra ones, such as EC2. Some other services, though, should be built on top of EC2, EBS, Lambda, S3, and etc, in which case Netflix and AWS teams use the same infra, yet Netflix internal services require much less oncall

No doubt Netflix is a much better run engineering org.

Its interesting that Netflix built their systems in a way that assumes the underlying platform is unstable. With Chaos Monkey and other systems they made sure things are resilient to flakey behaviour.

Reply on news.ycombinator.com