Drupal 6: The Performance and Developer Drupal release?

[The information in this post has been greatly updated - basically, Drupal 6 will enjoy even more of a performance advantage over Drupal 5 than is indicated here]

The more one learns about Drupal 6 the more there is to like. A while ago, right before Drupal 5 was released, I heard someone reference it being a 'developer's release' (think maybe it was Dries). That seems like an apt way to describe Drupal 6 these days. If the anti-CMS, ya-gotta-roll-your-own-or-it-stinks critics have a last breath left, this might be what snuffs it out once and for all. The list of new features, optimizations, and/or technologies introduced in Drupal 6 seems destined to light up the antennae of developers everywhere.

Performance, performance, performance

If there is any doubt about whether there will be performance improvements for Drupal 6, wonder no longer. Benchmarks, made using these standards, shows that the current, pre-beta, version of Drupal 6 performs 19.5% faster than Drupal 5.2 does when NO page caching is on. A healthy improvement by any standard, but even more so when one considers the following:

  • Drupal page caching is never active for logged in users, so any gains to non-cached speeds are pure gains for everyone that is logged in. This is a big deal when you consider that up till now it has taken 10x longer to serve authenticated/logged users than it does to server anonymous site visitors.
  • With the block cache patch which looks destined for core applied, Drupal 6 becomes 32% faster than Drupal 5.2 for authenticated users. Community site system admins rejoice.

Without quibbling over the details of specific numbers and/or benchmarking methods - one thing is very clear.

Speed is coming to Drupal.

  • Download benchmarks for Drupal 6, which includes benchmarks for block caching here.
  • Download benchmarks for Drupal 5-2 here.

10 August, 2007


As I said in my post about quibbling with exact methods or the numbers, and maybe I'm missing something else, but the only thing I didn't include in my post vs. the benchmarking standards is results which included the standard deviation. I did in fact look at the standard deviations of the tests and they told the same story (in fact they broke in favor of the improvements not the other way around). I did not include them because I was trying to keep this simple for everyone, but maybe I should have.

Seriously, all nitpicks aside, the bottom line is that for all the tests - across the board they showed a nearly identical percentage improvement in time per request AND the overall time it took for the test. While there are some things that numbers can hide - there's also somethings that numbers can't hide. In this case it's that Drupal is appreciably faster.

Wow, _that_ is patch advertising !

I do think you typed 51% instead of 15%, though ;-)

That is deduced by figuring out how much an improvement Drupal 6 benchmarks w/ block caching on is over benchmarks for authenticated users in Drupal 5. It really is a 51% improvement in the overall time.

Ah got it. That is D6 / block cache against D5.
But if I read the numbers correctly, 855ms in D6 against 1279ms in D5 is 'D6 takes 32% less time than D5'.

But 32% faster is still awesome :-)
Well, of course, the block-cache induced gain highly depend on the nature of the blocks on a page...

I guess I should brush up on my math. As you say it's still awesome, but decidedly different than I indicated. Will correct the post - thanx!