Hacker Read top | best | new | newcomments | leaders | about | bookmarklet login

Netflix is always the perfect example against this weird logic. They commited to AWS, and stuck with it even after Amazon itself is one of their prime competitors. They clearly decided that infra and devops is not their core competency and have absolved themselves of that responsibility and are ready to pay a premium for it.

On the other side we have people who keep patting themselves in their back about their Emacs setup being able to run a web server from a raspberry Pi.

We're at a place where it's fairly counter intuitive to continue arguing against cloud providers. I doubt azure or AWS are going to go down with a probability that should worry any business (and even then it should be recoverable). The only danger with cloud providers is that if your engineering team is not exactly smart you can rack up a million dollar bill for a dog walking app. But thars more about your ability to recruit disciplines engineers, no point blaming it on the cloud.

Look at Imgur. For the most part bootstrapped, scales extremely well, and if I'm not wrong, is very frugal, also runs off AWS.



view as:

> doubt azure or AWS are going to go down with a probability that should worry any business

Were you asleep when S3 East failed to the point that half of the internet was offline or did you just not get a notification?

https://aws.amazon.com/message/41926/

> The only danger with cloud providers is that if your engineering team is not exactly smart you can rack up a million dollar bill for a dog walking app

The public cloud has never been recession tested and these next few years may be the only way to do it. The model works as long as there are people willing to pay for Amazon and Azure will raise prices if the base load drops below a certain point.


Houses catch fire/flood/tornado too. The question is overall SLI, and multihoming, not has there ever been an outage.

And that half of the internet decided that it wasn’t worth it to have multi region redundancy. I’m not saying that was a bad decision. It all depends on how much they would lose in an outage vs. cost.

This also includes AWS because they stored the images for the dashboard in S3 East.

Just for being that guy: AFAIK Netflix has some caching devices that were pretty much home baked with FreeBSD and some other goodies. Can't remember the name, but I know they sit them in exchanges directly to ISPs. IIRC they were only using AWS for storing content with low frequency access.

They use AWS for much more than just “storing content”. You can read about all they do on AWS just by reading their blog and watching some of their ReInvent presentations.

Here's a list of Netflix blog posts about AWS.

https://netflixtechblog.com/tagged/aws


They call it OpenConnect: https://openconnect.netflix.com

If I remember correctly, they use S3 to store and distribute the master files, and other AWS services for the user interface and the like, but streaming itself is handled from their own infrastructure.


Legal | privacy