[Zope] www.zope.org's source?

chas panda@skinnyhippo.com
Thu, 21 Oct 1999 03:46:58 +0800


At 10:10 AM 10/19/1999 +1000, you wrote:
>+----[ Michel Pelletier ]---------------------------------------------
>|
>| The zope site is currently being hammered by ~50K hits per day, which
>| wouldn't be so bad if it weren't also our mailing list hub (which for
>| the Zope lists could take a whole machine by itself!).  Between Apache,
>| Zope, mailman and sendmail, the server averages a load of 2+.  The good
>| news is Zope is holding up to the abuse quite well (we are currently
>| looking for another server to take over the mail load).

I certainly hope that all of that load is actually the mailing lists.
A Pentium 166 with 128 MB RAM handles a million hits a day (10% pages
dynamically generated), no problem whatsoever (other than the ISP 
wanting more money) - average load less than 0.8  
Ezmlm and Qmail can handle considerable more traffic than the zope lists
in total without putting a strain on a P166 also.

What I'm getting at is this :

Are Python-based applications (Zope and mailman) are intrinsically 
more resource-intensive ?  If so, how and to what degree ?   I'm 
still intrigued by the fact that we had a site that was handling 2 
million requests/hits per day (albeit on a well-endowed 512 MB, PII 450 
machine) using nothing more than python CGI scripts (40% of *pages* were 
dynamic) but the mere act of running Zope/Zserver on it (zero load b/c 
it was in development mode and not made public yet) caused the kernel 
errors :
> file: table is full
> file: table is full
At the time, other FreeBSD users on this list mentioned seeing the same
and all of us just seem to up the MAXUSERS in the kernel. But doing
things without knowing why doesn't fill me with confidence, and we still
haven't put the zope section of the site live. 

Any clues ?

chas