I have performed a load test on on my website but don't understand if the result are good or bad.
Here is the server spec:
Server Linux/Ubuntu hardware (AMI on Amazon AWS) *-cpu:0 product: Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz size: 2500MHz width: 64 bits slot: CPU1 *-cpu:1 size: 2500MHz slot: CPU2 *-memory:0 capacity 3337MiB *-bank:0 description: DIMMRAM size: 15Gib width: 64 bits *-bank:1 description: DIMMRAM size: 15Gib width: 64 bits
Apache/2.4.18 PHP 5.6.18 (PHP-FPM)
Wordpress 4.5.2 multisite + Avada theme + W3 Total Cache + Yoast SEO
Test Load Settings Used:
The load scenario test is only done on 3 pages (home, product page 1, product page 2).
Google optimisation test:
My interpretation of the result is:
I have good server hardware, with less or more the latest version of all service (Apache, PHP) and software (Worpress). The response time of 50ms (AVG) is very good. The response time is stable even if it increases the number of users.
I'm not sure of my interpretation...
50 VUs, offered as free trial, by most load testing services is quite useless. Unless you have very heavy queries on your page, most servers (even shared hosting) will pass with flying colors. Wordpress is very scalable with caching on (as you have w3tc), so load test will definitely pass with 50 VUs or even at much higher number.
You need to try with higher number of VUs to see the impact and run it for longer time duration (at least another 3 min after reaching max VUs).
Also, it is ideal to monitor your server stats (memory usage, CPU usage, query times etc) while running the load test to locate the bottlenecks, if any, so you can optimize accordingly.
There's very little point in collecting metrics you don't understand. All this really tells you is that this system can comfortably handle the test load you tried it with (capacity). But a response time of 50 ms is the request response time which has very little to do with the page response time. Capacity is not the same thing as performance.
The Google rules are a bit more useful for identifying well defined problems with your performance, but as a simple set of rules their results should be treated with some caution. Really you should be looking at the time it takes to render a page. Webpagetest is great for this.
Unless you have a huge back catalogue of content to publish, and we're talking about thousands of pages here, and modest traffic volumes, this system is very badly specced. While you do need more memory to run mysql off remote storage, this looks like overkill.