Cloud save/cloud reliability megathread

[spkthed]

Okay, so this is a topic that I felt got a surprising amount of commentary. Be nice to un-clutter the patch threads and move it into here if there's any discussion left. Personally, I'm a fan. 

This I found highly interesting though, and wanted to discuss in more depth:
 

Quote

Forget it. Computer tech and services vendors are always vastly over-promising reliability and dependability. Phone companies had reached 99.999% uptime (before mobile phones). When "the cloud" is there, I will consider it reliable. 

 

Cloud computing is based on a distributed computing concepts, not linear computing concepts. For example, a phone company can't deal with a stadium full of people suddenly going to somewhere rural, and trying to use data. The whole system falls over. This is directly relevant because phone companies and 'The Cloud' use the same backbones and technological concepts. The Cloud *is* designed to deal with 'suddenly NYC has decided to relocate to small-town USA with 30 minutes of notice' so you can't really compare them.

There are ways using the cloud to hit reliability figures that traditional computing couldn't touch, but that's not what most people want or need, so that's not what most tooling is built for.

The Cloud *is* designed to deal with scaling on a very high level, which changes the entire premise. Instead of having linear workflows, building into an app state management and retries allows for the other end to deal with failover gracefully. This failover allows for a much higher reliability rate (https://wa.aws.amazon.com/wat.pillars.wa-pillars.en.html) than traditional datacenters, but inherently there's a tradeoff: security, reliability, efficiency, and cost all compete with each other, and every workflow has to find the right balance.