Re: Tux and load average

I have a box here, psyche, which dual boots
RH 8.0 and SuSE 8.1. It's a humble celeron
600 with 128 MB RAM and 2 18 GB ide drives.

At the moment it's running SuSE, and I can
test the relative performance of apache and
tux as follows using a small file to focus on
the server turnaround time.

(tux on port 80, apache on 8080)

ab -k -n 8192 -c 64 http://psyche/favicon.ico

and

ab -k -n 8192 -c 64 http://psyche:8080/favicon.ico

The apache test sends the load average up
over 10 immediately, and makes the desktop
jerky and sluggish.

The tux test doesn't seem to affect the desktop
at all, and the load average shows a barely
perceptible increase,  from e.g. 0.1 to 0.2

Tux also handles 8 times the number of requests
per second as apache, while hardly impacting
the server.

A crude test, but very clear.

Joe


Pierre POMES wrote:

>Hi all,
>
>I read that using tux can decrease load average. But according to the definition of
>load average (number of processes ready to run), this can be absolutly normal, since
>there is a part of tux running in the kernel (and not in one or some processes in the
>user space). This processing is not taken by the load average information.
>
>So, is there any way to compare the load of a machine running tux and a machine not
>running tux for the same web pages and the same network activity ? The "top" utility
>displays the % CPU idle, I suppose this number will be higher when running tux, and
>perhaps the % system CPU will be higher also with tux, but globally, the % CPU total
>activity should be lower with tux.
>
>Thanks for your help,
>
>  
>