Flarum included in Kinsta's latest PHP benchmarks

Darkle

I thought it was pretty cool to see Flarum finally included in Kinsta's benchmarks, it's the first time they do it, the only forum software by the way, Flarum is already getting noticed as the reference for PHP forums, you developers should be proud! 🥰

Talking about the results I think the most striking of all is that there is practically no difference between PHP versions, it is really quite similar to what happens in Laravel, although there is a little more difference with the new versions (except with 8.1, what happened there? 🤨).

Benchmark Results

Flarum 1.2.0 PHP 7.2 benchmark results: Unsupported 🚫
Flarum 1.2.0 PHP 7.3 benchmark results: 120.21 req/sec
Flarum 1.2.0 PHP 7.4 benchmark results: 122.06 req/sec 🏆
Flarum 1.2.0 PHP 8.0 benchmark results: 119.67 req/sec
Flarum 1.2.0 PHP 8.1 benchmark results: Unsupported 🚫

From a technical point of view, is there any reason why this happens particularly with Flarum/Laravel? Or are the big differences that do occur in other software simply due to a large number of factors that come together with the fact of using newer versions of PHP?

clarkwinkelmann

I'd need to do more tests or see a breakdown to say for sure, but I imagine Flarum's results will be impacted a lot by the database server since the homepage requires a number of SQL queries to be performed. Those queries would obviously take the same time to complete, no matter the PHP version.

I also wonder whether they visited Flarum before the initial query. Because if the CSS and javascript files aren't compiled, the first query will take a huge hit. If they just installed Flarum and didn't enable/disable any extension then it should be fine because the first redirect after installation should take care of generating those files.

I see in the article they say the Laravel test used few or none SQL query though, so not sure why it seems to correlate.

I'm also not sure why they marked 8.1 as unsupported. Officially it's supported, PHP display_errors just needs to be turned off.

EDIT: I see their screenshot is labelled "The tested Flarum page" and we see a user is connected. So this means they tested the logged in experience instead of guest (?) Also if they tested with an admin user the performance will be worse and even more database-tied than a regular user because more data is retrieved.