Panasonic Youth rob sanheim writes about software, business, ruby, music, stuff and things

Rails 1.2.1 is twice as slow as Rails 1.1.6?

This seems very strange to me - look at the performance between Rails 1.1.6 and Rails 1.2.1 in this performance comparison. The rails code is very simple - just set an instance variable in the controller and render it in the rhtml. How could this be so much slower in 1.2.1? The comments are mostly in Russian, so I’d love to hear comments from someone in core on this…

We found the same thing, our first look at the stats points to the view as consuming a lot of processing time. But we haven’t had enough time to digg on this issue.

This is a comparison between Rails 1.1.6 and 1.2.1 … not 1.6.1 (very distant future release).

Chris: You’re right, my latent dyslexia is showing. I’ve updated the post.

Now you need to update the topic’s title so the main page shows 1.1.6 too :)

Skan: It is — look up. The header is 1.1.6.

We actually haven’t noticed any speed difference between 1.1.6 and 1.2.1, probably because our app spends all its time in erb (whose speed has nothing to do with Rails proper) and memcached/mysql. What we have noticed is a massive increase in load. Big time.

As if Rails wasn’t already slow enough… am I the only one who thinks the core should focus more on cleaning things up (like ActiveRecord) instead of piling on more and more features?

Without a solid foundation to build on, things will get ugly.

In the article that posted the “2-4 times slower” analysis, there is a comment below that suggests the tests were run in development mode, not production mode, and that yes in 1.2+ development mode got a lot slower as it was reloading source a lot more. The blog author hasn’t rerun the tests unfortunately.

Nic: I dunno, look at comment #23 from the author there - he claims it was a misprint.

# Alrond Says:
January 30th, 2007 at 12:35 a.m.

2 Jules:
Thanks, it’s a misprint, I have really used production, already corrected :-)

Regardless, someone should attempt to duplicate those results with some performance tests to see what is really going on.

I just made a quick test, generated two rails installations, “rake rails:freeze:edge” them to TAG=rel_1-1-6 and rel_1-2-2, then “rake rails:update”, created the hello_controllerb.rb and index.rhtml files, configured nginx as described in http://docs.planetargon.com/Nginx_Configuration/.

Ran five mongrels in production mode and then the apache benchmark on another box: “ab2 -c 5 -n 1000 http://host:port/hello” a couple of times with each rails version.

Rails 1.1.6 gave around 130 req/sec and Rails 1.2.2 gave around 170 req/sec. Since I did not test for long I’m quite sure those numbers have a margin of error of around +-10, but even then 1.2.2 is still clearly faster with this embarrassingly simplistic benchmark.

Just my 0.02€

I just tried 1.2.1 and it seems to have about the same performance as 1.1.6 so 1.2.2 seems to have got some kind of performance improvements.

No, the rails 1.2.x branch is not twice as slow as the 1.1.x branch. See http://railsexpress.de/blog/articles/2007/04/01/rails-1-2-performance

Stefan: thanks for the detailed report. Good to see that the decrease is much smaller.