[Zope] Zope.org performance

[Ben Leslie]

Hi Michel!

On Mon, 18 Oct 1999, Michel Pelletier wrote:

> > -----Original Message-----
> > From: Sean Treadway [mailto:seant@factory.dk]
> > Sent: Monday, October 18, 1999 11:07 AM
> > To: Zope List
> > Subject: [Zope] Zope.org performance
> > 
> > 
> > Now, I am keen on Zope like nobody, but after going to www.zope.org
> > Friday at about 10:00 US/EST the site would not respond any 
> > faster than
> > 30 secs/hit?!?
> 
> Note that ~10am is one of the peak usage times for internet bandwidth.
>  
> > If the site's poor performance is explainable, then please, we want an
> > explanation.  If not, then what kind of options do we have if 
> > we plan to
> > deploy Zope in a similar, as I see it now, medium sized web 
> > site and why
> > aren't they used on www.zope.org?
> 
> The Zope site is currently under a huge strain due to the explosive
> popularity of Zope.
> 
> > This has become somewhat of a FAQ.  Are there any statistics 
> > about what
> > kind of setup Zope.org is running?  How many machines, pcgi 
> > or ZServer,
> > how many hits/day, mean response time, ZODB size, is it using 
> > ZEO, etc.?
> 
> It is _one_ little trooper of a PII 350Mhz.  It is running Apache 1.3.4
> through PCGI to ZServer running with six Zope threads.  It currently
> gets approx 50,000 hits per day.  The ZODB size, packed, is 150+ MB and
> slowly growing.  There is nothing really special about this hardware,
> just a straight up intel box with a SCSI drive and 256MB of memory.  The
> machine is hosted in Los Angeles, CA by Codeit computing on a 10Mbit
> ethernet.
> 
> The load average, right at this moment, is 6.28  (ouch!)
> 
> The machine serves as our web server, mail server, mail-list server
> (which sends over a hundred messages to over a thousand people a day,
> you do the math), and demo server.
> 
> We are right now in discussion with a major linux hardware vendor about
> getting a couple big boxes to take the strain.
> 
> The upshot of all this is that Zope has been running beautifully on the
> zope site under this huge strain; this is real world testing in action.
> No one anticipiated the massive surge in resources that happened over
> the past three months.  In June, we had 500K hits per month, in August,
> 1.1M.  This month looks like it's going to break all the records.


Along these same lines I was also after a bit of help.

Whilst my site is not get huge amounts of hits (yet :). I have been needing
to suck the content from using wget.

When I do however my system really starts to thrash badly and checking the 
top output I get this:

<snip>                        MEM                 CPU% MEM%
30200 www-data   6   0 60328  56M   872 R       0 13.2 45.5   1:40 python
30212 www-data   0   0 60328  56M   872 S       0  0.0 45.5   0:00 python
30213 www-data   4   0 60328  56M   872 R       0  9.0 45.5   3:11 python
30214 www-data   1   0 60328  56M   872 S       0 10.4 45.5   3:09 python
30215 www-data   8   0 60328  56M   872 R       0 18.7 45.5   3:01 python
30216 www-data  10   0 60328  56M   872 R       0 19.1 45.5   3:14 python
<snip>


My site is only medium size (~200 pages, 16MB), however as you see it seems
to use a fiar bit of memory, is it normal for it to use so much memory even
thought my data.fs is 16mb (packed).

The machine specs are:

PIII450, 128MB, IDE HDD



Any comments would be grateful, since I was planning on using Zope in a
much heavier load environment in the future.


Regards,

Ben Leslie