> Is it some sort of spying thing? Incompetence
It’s Microsoft we’re talking about so I feel like the latter is a given, and the former is a reasonable assumption.
Jokes aside, it does seem that somehow these hyperscaler cloud providers who tout the reliability of their multiple zones and global distribution sure do seem to always have some central dependency or single point of failure. AWS for example seems to have a global outage once every year or two at this point, or like us-east-1 problems always seem to affect other zones too. You’d think with their scale these clouds would be completely decentralized and redundant but apparently not.
Google basically is at this point, right? Not perfect, and obviously some things like DNS, Authn, and load balancing can’t be completely isolated. But overall, I think they put a ton of thought into removing software and hardware dependencies between regions, maintainence planning, and designing systems that don’t have outages correlated between regions.
The problem with decentralized redundancy is that it requires separate update schedules and process controls. This directly conflicts with the desire for convenience in rolling out new updates and code changes, etc. so the single point of failure tends to be deploying new code to everywhere at once.