Comparing Plone and Drupal speed evolutions

written by ccomb, on Oct 9, 2010 12:55:00 AM.

Ok, testing the home page of a CMS is generally a bad idea, it is not representative of the responsiveness of the application when people eventually use it. Anyway I did it. I had seen a few benchmarks comparing Plone 4 to previous versions, and comparing Plone with Joomla, Drupal or Wordpress. So I just wanted to check it by myself.

Comparing the homepages of successive Plone versions is perfectly relevant because their contents are very similar. Comparing the homepage of Plone and Drupal is not fair : a default Drupal without modules is pretty useless and offers almost nothing. A default Plone can be used without additional modules to create any website or intranet and offers a lot of features.

So I've installed all major versions of Plone since 2.5, and Drupal 5, 6 and 7 alpha7, without optimisations, and without caching.

The conclusions are:

  • Anyone running a Plone 2.5 or 3.x should seriously consider a migration to Plone 4. This will not always be easy, but it's possible. We, at Alter Way, are conducting a test consisting of a migration of an even older Plone 2.0 site with deprecated products, to 4.0.
  • Each new version of Plone is faster, while each new version of Drupal is slower.
  • Plone 4 (and even 4.0.1) is available now, while Drupal 7 release date is uncertain.

Please note that the interesting point of this curve is not the absolute number of req/s, but the slope of the curve.

comparing plone and drupal speed evolutions


  • Are each point of the graph obtained by one measure, or by several, as in Jon Stahl's post? And, in this last case, what are distributions dispersions of each (could be shown through boxplots for example)?

    (Please, let excuse this poor translation from french :-/ )

    Also, thanks for blogging this part of one of your two presentations!

    Comment by fero14041 — Nov 4, 2010 4:18:26 PM | #

  • It's the most basic bench with ab, and a thousand of requests. I've repeated the test a few times to be sure. I've not noted down the dispersion, because it is very low in both cases and has no interest here (it won't change the long term slope). Regarding my osdcfr presentation, it was unfortunately totally screwed by a display problem. I think I will blog about the aws.demo application because it is interesting to show.

    Comment by ccomb — Nov 5, 2010 12:47:36 AM | #