I have done some (theming etc) work for someone that runs their own Plone sites.
I noticed that my very basic setup with very small resources of the same sites, with just Apache in front and catching setup set to 'Without caching proxy' is MUCH faster than their 'real' setup with NGINX etc.
Out of curiosity (and without bothering my customer to much): is it possible for me to compare 'something' without having access to their server.
(In other words: I wonder if they have done something wrong with setting up their sites)
There are a lot of small things you can do that might slow it down. We discovered one production stack recently that had a setting to run the server in debug mode. That had slowed the site down by 2-3 times.
I got the (full) site and to theme/fix I put it 'on a server a had, with an xxx.medialog domain. My site 'did decedent', but the 'real site' is quite slow. Since I dont have access to the 'real site' server I wonder if it is possible to check why it is so much slower than mine.
Other things that might cause it? Swap. Badly configured cache param which is either top small or too large. Swap issues on varnish.
Anything that calls out to another process like ldap or custom code.
We had a issue once with a system lxml that ran much slower than a custom compiled one which affected diazo performance. No idea how you work that out externally. I guess you can user diazo.off=1 and see if that's still slow.
No, I have a small server (in Germany, actually), Ubuntu 16.4 if I remember right, which has a lots of Plone sites on it. I just installed with the Universal installer as a standalone, copied var-folder and eggs-section of buildout, added the apache rewrite rule (through webmin interface), enabled cache, imported the 'without caching proxy settings (from the control panel).
And it was much faster than their setup, which is a zeo setup, I think.
I checked with google speed test and 'pingdom' and the score/ranking was much higher.