[pmwiki-users] Hardware, Performance

Marcus prima at wordit.com
Mon May 5 10:50:59 CDT 2008


On Mon, 5 May 2008 15:17:49 +0200
PP Köln ZA 121 <cebius at polizei-koeln.de> wrote:
> Sometimes it takes a little bit to long loading pages.even only text
> pages need between 6 and 10 seconds for loading.

I wonder whether this could be a network issue too?

Try accessing the server on a direct cable link just to see whether the
server is on its knees under the strain, or it is a network bottleneck.
(I hope you have a few LAN ports). I would check that first because no
matter how powerful a server you have, it will all be lost unless the
network flows freely.


> There are 2500 clients wich could connect the IP of our wiki.
> There are 300 and more (counted by totalcounter) hits per day.

Only 300 hits spread over 8-10 hours? That's not much, is
it? Visitors are not downloading tons of GB and clogging the
network are they?

Unless 30-50 hit at the same time on this small server, it shouldn't
cause a 6-10 secs on a local network, should it?

 
> We do now have: Pentium III 400 MHZ, 512 RAM -
> what kind of server configuration is recommended?


I guess it depends on the server strain at a given time, the
performance distribution. If 80% of the visitors come into the
office at say 9am and hit the server at the same time it would be a
struggle for a PIII 400 MHz, 512MB. If the 6-10 secs is all day long
then something is wrong maybe elsewhere?

You don't need a very powerful machine, but going slightly higher,
you may need a box equivalent to PIII 1GHz with 1GB RAM. Maybe you have
an old one lying around in the company. If not, there are thin clients
the size of a book with those specs which also use only 10-20% of the
electricity of the old desktops, so they save money and are much
eco-friendlier.

You could also try running lighttpd (lighty) instead of Apache.
Especially on low power machines, lighty may help. AFAIK, everything
works with lighty. I am running it locally with pmwiki for testing.

In any case check the site with a statistical viewer like
Webalizer or Analog to see where the peaks are (both Open Source).


Marcus




More information about the pmwiki-users mailing list