[wplug] Any software that ...

Mark Dalrymple wplug at badgertronics.com
Fri Feb 1 17:38:41 EST 2002


> I am going to have say 3 different webservers up.  From a
> desktop I want to hit the servers quickly and see how they handle it.

The wild wonderful world of webserver benchmarking!  To do this
well, there's a couple of gotchas you would want to avoid:

* have a specific plan of what you want to test.  e.g.
  "I want to determine the number of 87K static pages per second each
   web server will serve."  or "I also want to determine the number of
   pages served per second given this php script hitting the postgresql
   database" or "given 15 pageloads a second, what kind of latency do
   I get for the first byte returned from the connections"

* use the more powerful machine(s) to abuse a weak machine.  If you
  go the other way around, the weaker machine might not have enough
  strength to really exercise the powerful one.

* make sure you're not just measuring the network bandwidth.  Even
  wimpy machines can saturate the wire if you have a low-speed network
  (and even maybe a high speed network.  Been awhile since I last did
  benchmarking) If you get nearly identical results from all of your
  webservers, this could be a cause.

  In that vein, try to do your testing on an isolated network.  You
  don't want someone downloading iso's of m0nk3yb0tPr0n on your
  network during one of your tests and skewing your results.

++Mark Dalrymple, markd at badgertronics.com.  http://badgertronics.com
  "scsi is not magic.  there are *fundamental technical reasons* why
   you have to sacrifice a young goat to your scsi chain every now and
   then"
    -- BitPoet



More information about the wplug mailing list