undefined | Better HN

0 pointscyberax6mo ago0 comments

AWS _had_ architected away from single-region failure modes. There are only a few services that are us-east-1 only in AWS (IAM and Route53, mostly), and even they are designed with static stability so that their control plane failure doesn't take down systems.

It's the rest of the world that has not. For a long time companies just ran everything in us-east-1 (e.g. Heroku), without even having an option to switch to another region.

0 comments

JCM96mo ago

So the control plane for DNS and the identity management system are tied to us-east-1 and we’re supposed to think that’s OK? Those seem like exactly the sorts of things that should NOT be reliant on only one region.

cyberaxOP6mo ago

It's worse than that. The entire DNS ultimately depends on literally one box with the signing key for the root zone.

You eventually get services that need to be global. IAM and DNS are such examples, they have to have a global endpoint because they apply to the global entities. AWS users are not regionalized, an AWS user can use the same key/role to access resources in multiple regions.

harshaw6mo ago

not quite true - there are some regions that have a different set of AWS users / credentials. I can't remember what this is called off the top of my head.

1 more reply

MichaelZuo6mo ago

The parent seems to be implying there is something in us-east-1 that could take down all the various regions?

j / k navigate · click thread line to collapse