Why is my system slow?

A decision tree over measurements you can actually take, not over theories. Each step names the reading that answers it; each ending names a constraint and how to confirm it before you act.

Step 1

Users say it is slow. Is it actually slow, and for whom?

How to actually check

Look at request duration percentiles split by endpoint over the last 24 hours, and compare against the same window last week. Do not start from a resource dashboard.