When autoscaling is worth the effort

This is one of those topics that looks small until it costs you something. Autoscaling is not free, and pretending otherwise leads to bad decisions.

Cloud work rewards teams who automate early and punishes teams who click through consoles. If it only works because one person remembers to do something, it does not work yet.

When it is worth it

Scaling rules need testing under real load, not in theory. It is worth being explicit about, because assumptions differ quietly. The failure mode is not doing it wrong, it is doing it once and assuming it stays done.

When it is not

If nothing downstream depends on it and nobody is complaining, it can wait. The reasoning matters more than the rule, because the rule has exceptions.

How to decide

A slow dependency will not scale with you. Getting it slightly wrong is survivable. Ignoring it entirely is not. It is the sort of thing that looks like polish right up until it costs you an enquiry.

How to tell if yours is fine

The bill is a design document: it tells you exactly what your architecture actually does. Three things worth confirming about autoscaling before you move on:

• Someone can say what the current setup is without going to look
• Scaling rules need testing under real load, not in theory — and you know whether that is true here
• There is a way to tell whether the last change to this helped

Worth checking on your own setup before it becomes someone else's problem to fix.