[Zope] Scalability issues

[Elliot Lee]

I recently set up a Zope+Squishdot site. If I run a loop like:

while true; do
	wget -o /dev/null -b --quiet -x http://developer.gnome.org:65348/gnome-news/
	usleep 250000
done

...the server eventually becomes unresponsive and does not recover.

The server is Zope 2.0.0b4, plus Squishdot 0.1.5a.localfixes and
Tinytables, running on Red Hat Linux 6.0, kernel 2.2.10ac8. Dual PII, lots
of RAM, blah blah. Depending on your setup, you may have to change the
usleep time to reproduce the problem.

This happens whether I run ZServer as the HTTP server, whether I run
Apache+pcgi without ZServer, or whether I run Apache+pcgi with ZServer.
I've tried changing the # of threads (-t in start script) to both 5 and
150, with no satisfactory results.

I'm not asking for a million requests per second, just that (a) a way to
handle overloads gracefully (b) a way to get more than a page request per
second out of a zope server running on a fairly beefy box.

Suggestions welcomed,
-- Elliot
Who me? I just wander from room to room.