I'm trying to run some performance optimizations, on a wordpress website running on 2 x 1gb NGINX, php7-fpm webservers, with two gluster file stores on 1GB servers, and a remote mySQL database ona 2gb server. The two web servers are load balanced behind a behind a nginx reverse proxy. All the connections are on private networking.
The site loads in about 1 second in Chrome or firefox, however google PageSpeed insights says my server response time ranges from 15- 18 seconds, the test doesn't even take that long. GTMetric has the total load time 2-4sec, pingdom, and webpagetests are inline with GTMetric.
I know it should still be faster than the 2-4 seconds, but does anyone know why google pagespeed would be much higher? Very concerned google internally uses the same tool to measure performance and is affecting my search results.
The largest bottle neck was glusterFS, disabling gluster, reduced the server response time from 12-15 secs to < 2.
Further adding micro caching with fastcgi cache on the web server and proxy caching on the load balancer (nginx reverse proxy) brought response times down to acceptable levels.
This does for now, but i still haven't solved the underlying issue with gluster, which will need to be addressed for scaling.