On 22:21 Wed 29 Dec , Ian Bicking wrote:
Hmm... the numbers seem kind of odd, hitting a distinct point when performance degrades. Have you tried any profiling? E.g., with hotshot:
http://python.org/doc/current/lib/module-hotshot.html
I couldn't figure out how to use hotshot...and when you search for "hotshot python run" you find that a lot of others can't figure it out either. do you know of another profiler...one that's documented?
Wow, I hadn't looked at the documentation. It really is horrible, isn't it? The profile module's documentation is a bit better, it seems. Maybe just something like:
import profile profile.run('serve_application(...)')
The results would probably be very interesting. Another interesting test might be a simple Webware app vs. a simple Twisted app (maybe just a static file for Twisted).
I ran the same tests against Apache2/WebKit and found that it scaled nicely (30-35% better performance than WSGIKit) as long as you don't factor in the hit the web server took from having to fork 20 or 30 more instances of Apache (at a loss of 10MB of RAM per instance).
Eric
------------------------------------------------------- The SF.Net email is sponsored by: Beat the post-holiday blues Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek. It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt _______________________________________________ Webware-devel mailing list Webware-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/webware-devel