lenya-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jann Forrer <jann.for...@id.unizh.ch>
Subject Re: DO NOT REPLY [Bug 32840] - replace sitetree script with one based on XMLHTTP
Date Fri, 08 Apr 2005 11:16:28 GMT
On Fri, 8 Apr 2005, Josias Thoeny wrote:

> <snip/>
>
> > Before, the sitetree was generated from the sitetree.xml file.
> > But now, the SiteTreeGenerator uses the SiteTree object, currently
> > a DefaultSiteTree, which holds the sitetree in a DOM.
> >
> > Generating the sitetree from the DOM means heavy load on the DOM
> > itself and since the getLabels() method of SiteTreeNodeImpl access
> > the DOM in an asynchronous manner, it is  most likely that the DOM
> > get corrupted (bug 33908).
>

I did apply the changes proposed by Felix in bug 34348 and I get rid of
the error messages and the disorganization of the sitetree due to above
mentioned problems :-) (which did occur when i put heavy load on a lenya
live site). Now I got zero error messages and the sitetree keeps in good
order :-)

> Concerning the performance, something else just came to my mind:
> It might make sense for the SiteTreeGenerator to implement the
> CacheableProcessingComponent interface. That way, its output could be
> cached by cocoon, which should increase the performance.
> The problem is how to know whether a cached result is still valid (i.e.
> how to implement the SourceValidity).
> Anyway, some profiling might be useful to see if this is really an
> issue.
>

As alread mentioned I did some load testing using JMeter. The Results are
probably of interest. I test a lenya live site access it directy via
servlet engine (tomcat) and the same site static exported (wget) access it
via apache.

My Computer:
linux           : Linux version 2.6.8 (SuSE 9.2)
model name      : Intel(R) Pentium(R) 4 CPU 3.20GHz
cache size      : 1024 KB

Testplan:
50 concurrent user access 10 times 2 pages (1000 Request)

Results (summarized):

Tomcat: 1.5 request / second
Apache: 100 request / sedond

There is certainly room to enhance the lenya/tomcat performance. Some
possibilites have already been discussed on this lis e.g.
mod_lenya, simple caching mechanism (as described by Michi Wechner some
times ago), using apache to access images ......

Jann

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lenya.apache.org
For additional commands, e-mail: dev-help@lenya.apache.org


Mime
View raw message