Mongrel users - Mar 2007 - [PIMP] Topfunky''s httperf PeepCode screencast (Zed A. Shaw)

Hi,

Thanks Zed - this is very interesting. One item in particular caught my 
eye: Does anyone on this list have any comments or validation that 
Rails 1.2.1 is 2-4 times as slow as Rails 1.1.6? Topfunky provided a 
link that purports what looks like really horrible performance and 
memory characteristics for Rails 1.2.1, even v. 1.1.6:

http://www.alrond.com/en/2007/jan/25/performance-test-of-6-leading-frameworks/

Did Rails 1.2.2 fix anything in this regard? (I didn''t see any mention 
of performance improvements in the dev notes).

There were some nice feature upgrades in 1.2.1 that I''m using, but it 
would be great to know that performance is now a real issue (and 
rolling back would fix it).

Any information?

Thanks,

Steve

>Message: 4
>Date: Tue, 6 Mar 2007 11:08:24 -0800
>From: "Zed A. Shaw" <zedshaw at zedshaw.com>
>Subject: [Mongrel] [PIMP] Topfunky''s httperf PeepCode screencast
>To: mongrel-users at rubyforge.org
>Message-ID: <20070306110824.8a3e1d54.zedshaw at zedshaw.com>
>Content-Type: text/plain; charset=US-ASCII
>
>Hey Everyone,
>
>Topfunky (Geoffry) has created a great PeepCode screencast on using
>httperf to performance measure your web applications:
>
>http://nubyonrails.topfunky.com/articles/2007/03/05/peepcode-page-caching-and-httperf
>
>http://peepcode.com/products/benchmarking-with-httperf
>
>Topfunky put a ton of work into this and really got it right.  The
>attention to detail is very good and he researched the subject
>thoroughly.  I also spent time reviewing it so that he''d get the
>statistics right, but he did a lot more than I expected.
>
>The main thing that you''ll get from his screencast is the actual
>*process* of doing an analysis and then determining if there''s
>differences.  He shows graphs, takes notes, covers basic statistics,
>and shows you how to figure out how fast your Mongrel setup can go.
>It''s much better watching someone else do it than reading a 
>description.
>
>Everyone who constantly asks about how to performance measure their 
>web
>apps should grab it and go through it.
>
>Why am I pimping it so hard?  First, he''s giving me a cut of the
dough
>and a brotha''s gotta eat.  :-)  Second, it is probably the first
>description of performance measurement and analysis I''ve seen that
>actually gets it right.
>
>Thanks for your time, and enjoy the day.
>
>--
>Zed A. Shaw, MUDCRAP-CE Master Black Belt Sifu
>http://www.zedshaw.com/
>http://www.awprofessional.com/title/0321483502 -- The Mongrel Book
>http://mongrel.rubyforge.org/
>

Hmm, reading that I am growing more suspicious. My ram-related woes  
appeared when Rails 1.2 hit the stage. But I also saw a hefty  
increase in traffic so was quick to blame it on that.


Op 6-mrt-2007, om 20:12 heeft Steve Midgley het volgende geschreven:
> Hi,
>
> Thanks Zed - this is very interesting. One item in particular  
> caught my
> eye: Does anyone on this list have any comments or validation that
> Rails 1.2.1 is 2-4 times as slow as Rails 1.1.6? Topfunky provided a
> link that purports what looks like really horrible performance and
> memory characteristics for Rails 1.2.1, even v. 1.1.6:
>
> http://www.alrond.com/en/2007/jan/25/performance-test-of-6-leading- 
> frameworks/
>
> Did Rails 1.2.2 fix anything in this regard? (I didn''t see any
mention
> of performance improvements in the dev notes).
>
> There were some nice feature upgrades in 1.2.1 that I''m using, but
it
> would be great to know that performance is now a real issue (and
> rolling back would fix it).
>
> Any information?
>
> Thanks,
>
> Steve
>
>
>> Message: 4
>> Date: Tue, 6 Mar 2007 11:08:24 -0800
>> From: "Zed A. Shaw" <zedshaw at zedshaw.com>
>> Subject: [Mongrel] [PIMP] Topfunky''s httperf PeepCode
screencast
>> To: mongrel-users at rubyforge.org
>> Message-ID: <20070306110824.8a3e1d54.zedshaw at zedshaw.com>
>> Content-Type: text/plain; charset=US-ASCII
>>
>> Hey Everyone,
>>
>> Topfunky (Geoffry) has created a great PeepCode screencast on using
>> httperf to performance measure your web applications:
>>
>> http://nubyonrails.topfunky.com/articles/2007/03/05/peepcode-page- 
>> caching-and-httperf
>>
>> http://peepcode.com/products/benchmarking-with-httperf
>>
>> Topfunky put a ton of work into this and really got it right.  The
>> attention to detail is very good and he researched the subject
>> thoroughly.  I also spent time reviewing it so that he''d get
the
>> statistics right, but he did a lot more than I expected.
>>
>> The main thing that you''ll get from his screencast is the
actual
>> *process* of doing an analysis and then determining if there''s
>> differences.  He shows graphs, takes notes, covers basic statistics,
>> and shows you how to figure out how fast your Mongrel setup can go.
>> It''s much better watching someone else do it than reading a
>> description.
>>
>> Everyone who constantly asks about how to performance measure their
>> web
>> apps should grab it and go through it.
>>
>> Why am I pimping it so hard?  First, he''s giving me a cut of
the
>> dough
>> and a brotha''s gotta eat.  :-)  Second, it is probably the
first
>> description of performance measurement and analysis I''ve seen
that
>> actually gets it right.
>>
>> Thanks for your time, and enjoy the day.
>>
>> --
>> Zed A. Shaw, MUDCRAP-CE Master Black Belt Sifu
>> http://www.zedshaw.com/
>> http://www.awprofessional.com/title/0321483502 -- The Mongrel Book
>> http://mongrel.rubyforge.org/
>>
>
> _______________________________________________
> Mongrel-users mailing list
> Mongrel-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/mongrel-users
-- 
www.gomagazine.nl +31643904460 pobox 51059 nl-1007eb amsterdam

On Tue, 06 Mar 2007 11:12:00 -0800
Steve Midgley <public at misuse.org> wrote:
> Thanks Zed - this is very interesting. One item in particular caught
> my eye: Does anyone on this list have any comments or validation that 
> Rails 1.2.1 is 2-4 times as slow as Rails 1.1.6? Topfunky provided a 
> link that purports what looks like really horrible performance and 
> memory characteristics for Rails 1.2.1, even v. 1.1.6:
> 
>
http://www.alrond.com/en/2007/jan/25/performance-test-of-6-leading-frameworks/
> 
> Did Rails 1.2.2 fix anything in this regard? (I didn''t see any
> mention of performance improvements in the dev notes).
> 
> There were some nice feature upgrades in 1.2.1 that I''m using, but
it
> would be great to know that performance is now a real issue (and 
> rolling back would fix it).
I have been doing my own investigations into this issue, in particular,
I''m looking for at least one cause of an unacceptable slowdown that I
have been encountering (I will now go back and redo the tests under
1.1.6 for comparison).  In particular I''ve been observing *really poor*
performance with ActiveRecord.  In particular, when I run with the
profiler I''m seeing a stunning number of calls to the inflector
methods.

Here "my" deal:

I have a class that only inherits from ActiveRecord so I get some SQL
logging.  It actually is using the driver directly (in this case
unixODBC) because I am calling a stored procedure that returns multiple
result sets (something ActiveRecord cannot do). So I have a "find"
method that returns a simple collection of hashes, nothing more.  I
benchmarked 100 calls and found that the inflector methods were the
overall most time consuming part of the effort with over 40K calls!
Now, the stuff returned form the ''find'' call are not even
derived from
ActiveRecord so I would expect a total of 0 (zero) such calls.

The best I can tell is that there must be a whole lotta new after
filters in 1.2.x doing lots of Rails magic.

Still investigating.

Jim Powers

On Tue, 6 Mar 2007 21:13:06 +0100
joost baaij <joost at spacebabies.nl> wrote:
> Hmm, reading that I am growing more suspicious. My ram-related woes  
> appeared when Rails 1.2 hit the stage. But I also saw a hefty  
> increase in traffic so was quick to blame it on that.
I haven''t investigated it yet but you should probably test it out and
double check you''ve got fastthread in there.  If you run mongrel_rails
with -B you can get a log/mongrel_debug/objects.log file that will help
you track down the memory problems.

Let me know what you find.

-- 
Zed A. Shaw, MUDCRAP-CE Master Black Belt Sifu
http://www.zedshaw.com/
http://www.awprofessional.com/title/0321483502 -- The Mongrel Book
http://mongrel.rubyforge.org/

Allrighty, mongrel -B is running and producing the log.
Do you have any pointers on how to interpret the log?

on fastthread:
joost at ded022/home/websites/gomagazine$ gem list fastthread

*** LOCAL GEMS ***

fastthread (0.6.4.1)
     Optimized replacement for thread.rb primitives




Op 7-mrt-2007, om 19:38 heeft Zed A. Shaw het volgende geschreven:
> On Tue, 6 Mar 2007 21:13:06 +0100
> joost baaij <joost at spacebabies.nl> wrote:
>
>> Hmm, reading that I am growing more suspicious. My ram-related woes
>> appeared when Rails 1.2 hit the stage. But I also saw a hefty
>> increase in traffic so was quick to blame it on that.
>
> I haven''t investigated it yet but you should probably test it out
and
> double check you''ve got fastthread in there.  If you run
mongrel_rails
> with -B you can get a log/mongrel_debug/objects.log file that will  
> help
> you track down the memory problems.
>
> Let me know what you find.
>
> -- 
> Zed A. Shaw, MUDCRAP-CE Master Black Belt Sifu
> http://www.zedshaw.com/
> http://www.awprofessional.com/title/0321483502 -- The Mongrel Book
> http://mongrel.rubyforge.org/
> _______________________________________________
> Mongrel-users mailing list
> Mongrel-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/mongrel-users
-- 
www.gomagazine.nl +31643904460 pobox 51059 nl-1007eb amsterdam


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2423 bytes
Desc: not available
Url :
http://rubyforge.org/pipermail/mongrel-users/attachments/20070307/730265ec/attachment.bin