Design Implications For Going Serverless

In some cases it is cost effective, others not so much, but when it is we should consider what the adjusted design metrics should be. Going serverless doesn't mean there are no servers it just means we have left that responsibility (security/OS updates ... etc) to the owners of the infrastrure. That being said there are a few ways this changes the way we design, for example how do we communicate errors to both users and administrators?