[Zope-CMF] Plong/CMF lack of speed?

[alan runyan]

if you are going to test use ab, or microsofts web stress tool.
And I would like for you guys to count how many requests are
done on an initial Plone request.  We have ~10 CSS, and a
few external .JS files, thats not including images (transforms, etc).
On subsequent pages are any files being 'acquired' accidentally?
things like this will kill performance <img src=3D"folder.gif" />
which will give you a /folder.gif when you request the root or
/subfolder/folder.gif when you request 'subfolder' - same image
downloaded two times.

Plone interface will most likely never be a speed demon.  It is
designed for flexibility.  PageTemplates are slower - especially
if you have lots of metal calls.  Also its questionable if FSObjects
participate in the ZODB caching machinery (sorry its early
in the morning, is this correct? or has it been fixed?).  You can
use Richard Jones Call Profiler (slick piece of software).

You should create your own delivery
skin and test that, http://emerging.clearnoodle.com/ or http://moz.zope.o=
rg/
are good examples of this.

and since ZOPE takes a huge hit on static files being served.
I dont think its over-the-top for people to enable mod_proxy
for a plone site.  If we have 5 .css and zope is set for 3 threads
you can see upfront we are going to be waiting on ZOPE. (testing
with AB will rid you of this)

Testing the header seperate from the entire page can provide some
insight.

http://yoursite/header
http://yoursite/frontpage (go directly to the Document object)
http://yoursite/index_html (a page template that addes items to left/righ=
t
navigation slot and calls frontpage/CookedBody )

*just in header* there are 15 files that come along with the page.
10 css
1 javascript
4 images

this is why subsequent page's *feel* much faster - client side caching.

but ab will only request 1 file and ignore the support files
(.css,.jpg,etc).
on my Celeron 300Mhz - I just ran ab, http://mysite/frontpage and I
got ~2.36 req/sec. That isnt great.. but I dont think it sucks.
I just ab'd http://mysite/header and got ~4.52 req/sec.

~runyaga
p.s. I think the problem lies in how complex our setup is.  I believe
simpler PageTemplates, less macro inclusions, and more long running
functions that generate content would result in speedier results.

----- Original Message -----
From: "Gr=E9goire Weber" <gregoire.weber@switzerland.org>
To: "Jan Ulrich Hasecke" <janulrich.hasecke@web.de>
Cc: <zope-cmf@zope.org>
Sent: Monday, September 09, 2002 9:58 AM
Subject: Re: [Zope-CMF] Plong/CMF lack of speed?


> Hi Jan,
>
> it depends on the definition of 'tool' :-)
>
> My tool was Mozilla (it shows page loading time) and a
> chronometer for lynx (so times for lynx are probably
> about 0.4secs better than in the list).
>
> You can do your own automated measurements with Pythons
> httplib module combined with the time module.
>
> httplib examples:
>
>    http://www.python.org/doc/current/lib/httplib-examples.html
>
> time module:
>
>    http://www.python.org/doc/current/lib/module-time.html
>
> Greetings, Greg
>
> At 16:29 09.09.2002 +0200, Jan Ulrich Hasecke wrote:
> >Gr=E9goire Weber <gregoire.weber@switzerland.org> writes:
> >
> >>    (1) Mozilla from zope on localhost, Javascript on:  7.0 seconds
> >
> >Do you have a tool for measurement?
> >
> >Ciao!
> >juh
> >--
> >Der dornige Zweig der Abderiten-Familie
> >http://www.sudelbuch.de/2001/20010221.html
>
> _____________________________________
> Gr=E9goire Weber
> mailto:gregoire.weber@switzerland.org
>
>
> _______________________________________________
> Zope-CMF maillist  -  Zope-CMF@zope.org
> http://lists.zope.org/mailman/listinfo/zope-cmf
>
> See http://collector.zope.org/CMF for bug reports and feature requests
>