We optimized a LOT!
Shai Almog

First of all sorry if I wasn’t clear, I didn’t mean you didn’t optimize at all, just not efficiently enough. I thought about how would be our project right now without using the aforementioned solutions — like you did — , and well it would be at a much worse place… but I never expected Google to give solutions to my implementation problems. What I failed — ant still fail — to grasp regarding your stance is that you keep telling how bad it is that they didn’t show you the exact place of your issues. Well how using IaaS wil do that? If you start experiencing performance and/or cost issues you will have to investigate it yourself anyway. Just because an “I” is replaced with a “P” I don’t magically expect everything to be provided automatically by the platform. My point is trying to be that if something isn’t given, but it can be done, well just do it.
I can’t say GAE is perfect — that was never my goal — but I do see a pattern of getting better and better over the years — from my point of view — , and I prefer it much more compared to for example application servers.

BTW the answer to your last question is obviously no. If someone says everything is perfect, he is either naive or lies :) All I say is that we were able to find and fix a lot of issues easily (in a cost effective manner) with the solutions I told you about. Right now there isn’t any significant item in the cost analysis, that we don’t think is valid. Guessing that something might cost a lot or knowing that something does makes a lot of difference in optimization costs. I can’t really say exact percentage how much decrease we achieved — as it was over the years and usage obviously changed — , but I know it’s a lot, and we never sweated over it.