Posted by ken hawkins on August 11, 2009 at 11:47am
I recently saw someone post in regards to seeing a significant speed boost by switching to PressFlow's "enhanced/streamlined" version of Drupal.
However some googling and reading revealed no benchmarks on how much faster it is on average. Does anyone know how much of a speed gain we're talking about?
For comparison: The number I've seen tossed around for PHP accelerators is 15% for more complex sites.
Comments
Pressflow will help if you
Pressflow will help if you have multiple MySQL databases that replicate in some way. It won't change your world on single server installations as far as I know. It does do quite a few things, one of them being marking some queries as runnable on the slave MySQL servers e.t.c. I think it also likes to be installed on innodb and removes some silly lock table statements. Although I could be getting very confused now.
When you have multiple db servers it will make a difference.
It is still a killer distro to have though. We use it at Economist.com.
Full Fat Things ( http://fullfatthings.com ), my Drupal consultancy that makes sites fast.
The query lock
The query lock optimizations, optimized SQL, lazy session creation, proper caching headers, and native JSON will all help even on a single machine.
Keeping the wolf from the door
We had David come in and do some consulting with us on site performance, and we've switched to using Pressflow's optimized Drupal on our newspaper servers.
I don't think of Pressflow as our primary tool to make pages deliver faster. Most of our experience-killing problems have been our own fault. Our biggest page-performance increases have come from attention to detail in the use of HTML, CSS, sprites and other tools focused on object size and client-side rendering.
Where Pressflow really helped us is in avoiding having the system dog down under heavy load. The proper caching headers can't be overstated in terms of importance, particularly if you're fronting Drupal with Squid/Squirm. Since we did the tuning we've gone from one or two crisis situations a day to weeks without any performance trouble, and we're not (yet?) running a replicated database.
Memcache/Authcache
Having a multi-instance memcache/authcache backend will also give additional performance gains for both unauthenticated and authenticated users.
-- Michael
.. but any benchmarks?
Thanks for the run-down of the benefits of PressFlow - it definately looks like the solution to current scalability issues with Drupal 6.
From what I have read on other posts, many of the improvements of PressFlow 6 have also been added to Drupal 7, so I will be really interested in seeing what the guys at Four Kitchens will offer on top for their release of PressFlow 7. I think it's a really exciting branch of Drupal and worth keeping an eye on.
.. but as Ken Hawkins asked, are there any benchmarks available to show these improvements in performance: Drupal 6 vs PressFlow 6?
Have you seen this?
http://groups.drupal.org/node/25617
Project Mercury Benchmarks: 2000+ Requests Per Second!
Peace, Compassion, Prosperity
Any caveats in porting an existing Drupal site to PressFlow?
Is there a list of all requirements and implementation pitfalls for porting an existing D6 site to PressFlow?
(we're currently using EC2 (Centos) and want to squeeze more performance out of the site so want to go down the PressFlow+Varnish+Memcached route.
Are there any migration horror stories we should be aware of?
Or better yet, any "shortcuts"?
Shouldn't be too bad
I've moved a number of smaller sites and as long as the tech requirements are amble (mainly php 5.2+) it's pretty much a drop in replacement.
Obviously on larger site there may be more considerations if there's lots of custom code.
I've yet to hear of a Pressflow horror story, likely as it seems to mainly be a streamlined Drupal install designed to play nice with external caching.