Yeah, AWS is still one of the best out there, but this outage is a good reminder that no cloud is bulletproof. Trust them, sure, just don’t build like they’ll never fail. Always plan for downtime, use multiple regions, and have a backup plan. Reliability in the cloud is a shared job, not something you can fully outsource.
As far as cloud providers go - sure. We've had numerous "wats" with AWS, but I don't think it would have been better at, say, Azure.
I do think we'd have avoided a lot of trouble by investing in our own servers rather than paying more in the long run for a cloud provider which will always come with more risk (more moving parts), and _by definition_ has to cost more than DIY in the long run because otherwise cloud providers wouldn't make money.
The one problem I have is the over-reliance on the big dogs - so an AWS DNS oopsie (which aren't as rare as I'd like - unrare enough that there are memes about it) causes so many things to break on the internet. Not just my work stuff - I couldn't play Helldivers either because GameGuard apparently relies on AWS too. Slack has issues. Even Tailscale, which is supposed to not rely on AWS, apparently does in some respect - when AWS was being dodgy, so was Tailscale.
There's a huge warning buried in this key-man reliance.
Yeah, AWS is still one of the best out there, but this outage is a good reminder that no cloud is bulletproof. Trust them, sure, just don’t build like they’ll never fail. Always plan for downtime, use multiple regions, and have a backup plan. Reliability in the cloud is a shared job, not something you can fully outsource.
People just need to not build in us-east-1 and they’ll be fine
As far as cloud providers go - sure. We've had numerous "wats" with AWS, but I don't think it would have been better at, say, Azure.
I do think we'd have avoided a lot of trouble by investing in our own servers rather than paying more in the long run for a cloud provider which will always come with more risk (more moving parts), and _by definition_ has to cost more than DIY in the long run because otherwise cloud providers wouldn't make money.
The one problem I have is the over-reliance on the big dogs - so an AWS DNS oopsie (which aren't as rare as I'd like - unrare enough that there are memes about it) causes so many things to break on the internet. Not just my work stuff - I couldn't play Helldivers either because GameGuard apparently relies on AWS too. Slack has issues. Even Tailscale, which is supposed to not rely on AWS, apparently does in some respect - when AWS was being dodgy, so was Tailscale.
There's a huge warning buried in this key-man reliance.