Review Performance: SSD VPS 3 Plesk + WordPress + FastCGI Apache

Review with SSD VPS 3 :
3GB ram, 60GB SSD disk, 3vCPU, 100 Mbps network speed, KVM Virtualisation

Specs :

  • Latest Plesk
  • PHP-FPM by Nginx
  • PHP 7.4 stable
  • memory_limit 128 MB
  • max_execution_time 30 secs
  • max_input_time 60secs
  • post_max_size 8M
  • upload_max_size 2M
  • opcache.enable On
  • FakerPress plugin for fake content (15 post with random picture)

+ Test with 20 VU +

First test using 20 VU (Virtual Users) within 1 minutes.

Summary (k6.io)

  • Server tested from Singapore.
  • Total handled request is about 802 (higher is better).
  • Highest peak request is about 18.67 (from 20).
  • Total HTTP failures is 0 (lower is  better) which is GOOD.
  • Average response time is 221 ms (lower is better) with highest reponse time is about 591 ms (see blue line).
  • First test showing 2 spikes (a huge spike) that showing latency become high. It explains how FastCGI work that can’t maintain the latency become ‘almost flat’or stable even under small visitor simulation.

Page load time (Gtmetrix)

  • Server tested from Canada
  • Gtmetrix grade showing B with 81% performance and 88% structure (a bit downgrade from php-fpm nginx and php-fpm apache)
  • Fully loaded time within 2.3 s
  • Largest contentful to be loaded is 1.7 s

+ Test with 50 VU +

Second test using 50 VU (Virtual Users) within 1 minutes.

Summary (k6.io)

  • Server tested from Singapore.
  • Total handled request is about 1471 (higher is better).
  • Highest peak request is about 29.67 (from 50).
  • Total HTTP failures is 0 (lower is  better) which is GOOD.
  • Average response time is 667 ms (lower is better) with highest reponse time is about 1818 ms (see blue line).
  • Second test also showing spike result of latency (blue line) under 50 visitors within 1 minutes. The graph showing FastCGI performance has a spike at first and the end of request.

Conclusion

FastCGI might be a good choise for low-end specs and very low visitor with low-average image website. It was explained by the result of both load impact test that at first and the end of request, FastCGI showing spike which affected to its latency. For a high traffic, FastCGI might not be a good choise.