Speed matters – Episode 2

About one year ago, i wrote about the compared performance of WPML, qTranslate and Polylang. Of course I cannot test every multilingual plugins as there are now quite a lot on the market and thus decided at that time to focus on the two most widely used plugins.

However, over the past year, I read a lot of blog posts praising the benefits in performance (speed) of multilingual plugins relying on WordPress multisite feature. Here are some quotes found on the internet about Multilingual Press:

Use WordPress’ built-in core features to create networks with multiple languages without losing any performance.

Performance. This plugin uses WordPress’ built-in core features. For large WordPress sites is the best option.

One pro is the performance: single site plugins must use custom post types, they have to filter almost any database query to get the correct languages. This is slow, much slower than a simple switch_to_blog() behind the scenes.

It doesn’t harm your website performance.

But unfortunately, none of these assertions come with facts. I am curious and wanted to know more, and so finally decided to compare the free version of Multilingal Press 2.1.2 to Polylang 1.6.3.

I used exactly the same test protocol as last year except for the version of WordPress (4.1 instead of 3.7.1). Note that WordPress alone and Polylang are tested on single site installations, whereas Multilingual Press is obviously installed on a multisite installation.

Here are the first results when querying the English version (no .mo file is loaded). Figures reported are average total time per request in ms:

Non localized (en_US) WP alone Polylang Multilingual Press
Home page 426 492 520
Page 402 477 505
Monthly archive 433 498 513
Category archive 423 499 532
Post 421 502 520
Average 421 494 518

Below are the results when querying the French version (the French .mo files are loaded including for the site with no plugin for which the language is set to French in the WordPress general settings):

Localized (fr_FR) WP alone Polylang Multilingual Press
Home page 475 545 567
Page 449 526 552
Monthly archive 477 547 559
Category archive 471 541 584
Post 464 544 556
Average 467 541 564

And here is a simple table to summarize the performance impact of Multilingual Press and Polylang compared to a fresh WordPress (single site) installation.

Polylang  Multilingual Press
Added time 16%  22%


The first conclusion of this test is that the performance of WordPress itself has decreased quite a lot in one year! Just compare the figures with those obtained one year ago for a WordPress installation without any plugin. Yes I remade the tests with WordPress 3.7.1 to check!

The second conclusion is that the performance of Polylang has neither improved, nor decreased in one year.

The third conclusion is that Mutilingual Press is certainly more performant than qTranslate or WPML, but relying on the multisite feature of WordPress does not give it a significant advantage. And in fact, it seems that Polylang takes more advantages of “builtin core features” – although not the same – to get a significantly better performance.

2 thoughts on “Speed matters – Episode 2

  1. RavanH

    Hi Frédéric, interesting results!

    I use Polylang a lot on sites that are running on WordPress Multisite so I wonder how much of this (be it small) difference is due to Multisite. If you have your test setup still there, maybe convert the install with Polylang to Multisite too and run the test (on Polylang) again? It might just add those extra milliseconds…

    1. Frédéric Post author

      Hi Rolf,

      I did not make this exact test (Polylang on a multisite), but yes you are right, part of the time penalty in Multilingual Press is due to the multisite installation. I have evaluated this as I just had to de-activate Multilingual Press. The same site (without any plugins, then) is in average 4% slower on a multisite installation than on a single site installation. So probably Polylang on a multisite installation is very close to Multilingual Press.

      To be honest, I was a bit surprised myself by these results. I you had questionned me a couple of months ago, I would have certainly answered that probably Multilingual Press and other multisite based plugins could be a bit faster than Polylang. Just because all of these filters in Polylang…

      But then I have tested Multilingual Press and have thought deeper to this. OK Polylang, WPML and similar plugins have to use a lot of filters to separate the content. But multisite based plugins have to create connections. And this seems not to be less expensive…

      To go further I had a look at the number of queries (reported by the debug bar plugin) on the homepage of my test site:
      * single site without plugins : 22
      * Polylang: 25
      * Multisite without plugins : 32
      * Multilingual Press : 41
      It illustrates clearly how connecting the content is not free.

Comments are closed.