Dennis, This is in direct correlation with the check interval suggestion I made. In your search, you will find that on SMTP systems, tweaking the check interval can help things, precisely because of the GIL. I set it to zero because I had the spare CPU power, and trying to fully understand/benchmark python's behavior on SMTP seemed like too much work :P J.F. -----Original Message----- From: zope-bounces@zope.org [mailto:zope-bounces@zope.org]On Behalf Of Dennis Allison Sent: Tuesday, March 23, 2004 10:39 AM To: Chris Withers Cc: zope@zope.org Subject: Re: [Zope] strange performance problem Thanks. I'll review the archives. -d On Tue, 23 Mar 2004, Chris Withers wrote:
Dennis Allison wrote:
While it's possible, the scheduler is pretty good at maintaining processor affinity without intervention. I have *never* seen anything which would suggest that is a problem. This is happening on an unloaded system when processor affinity issues should be hard to trigger.
Okay, but I have lots of memories of reading about problems with Python's Global Interpretter Lock and lightly loaded multi-processor machines.
The mailinglist archives may give more of a clue.
One thing you could try is to get a ZEO storage server up and running, then run up 4 or 5 ZEO clients, but each with only one thread. Then you just need to figure out how to round-robin them so they all appear through one port...
cheers,
Chris
-- Simplistix - Content Management, Zope & Python Consulting - http://www.simplistix.co.uk
_______________________________________________ Zope maillist - Zope@zope.org http://mail.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://mail.zope.org/mailman/listinfo/zope-announce http://mail.zope.org/mailman/listinfo/zope-dev )