On May 20, 8:35 am, cd34 <[email protected]> wrote:
> On May 19, 5:12 pm, Graham Dumpleton <[email protected]>
> wrote:
>
> > Paste server would likely preload TG. In Apache/mod_wsgi it would lazy
> > load it on demand when first request to a process arrives. How did you
> > ensure your tests against Apache/mod_wsgi weren't counting the cost of
> > loading TG?
>
> 3 runs of apachebench, fastest time recorded.  For mpm-prefork, this
> eliminated any apache child creation issues and the wsgi environment
> should have been created and preloaded.  During ab, it was obvious
> when a preload was hit and that result was thrown out.  mpm-worker
> seemed more consistent from first through third run.  Best case of the
> 'worst case' scenario seemed to be a reasonable assessment.
>
> I am sure there are settings in TG2 that will affect this.  Perhaps I
> don't have all of the debugging messages turned off properly or have
> missed some of the options, but, I don't think that is going to
> materially affect the results.
>
> I am somewhat sure that the 2nd and 3rd runs would have had things
> preloaded since I was calling the same page over and over again.  The
> fact that a simple 'Hello World' wsgi app handled over 2000 requests
> per second suggests that wsgi is working.
>
> At this point, I need to profile TG2, SQLAlchemy and many of the other
> modules.  SQLAlchemy backreferences appear to induce a hit.  Looking
> at the queries that hit mysql, I'm sure there are some improvements
> that can be made to my auth handler to remove half a dozen queries and
> reference links, but, my initial goal was to figure out where the
> bottlenecks were and to figure out what other people were doing for
> deployment.
>
> Right now, it seems like apache2|nginx/ssl/mod_wsgi is what I need for
> this project.  For my future project, I've got to do quite a bit more
> investigation to see what sort of resources I'm going to need.

Some related reading for your then:

  blog.dscpl.com.au/2009/03/load-spikes-and-excessive-memory-
usage.html
  http://blog.dscpl.com.au/2009/05/blocking-requests-and-nginx-version-of.html

Sorry about the earlier empty followup. :-)

Graham
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"TurboGears" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/turbogears?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to