Not exactly a technical but more philosophical question.
Currently I am dealing in more reactive mode vs. proactive. When CPU or memory hits the limit we scale our server farm with more servers.
What I would like to do is to predict and add servers before we hit the limit and loose customers. So we can avoid a downtime and things will be smooth and transparent.
Anyone has a good knowledge about that? Advice? How to predict and scale in predictive manner vs. reactive?