Performance Stats

[Arran Cudbard-Bell]

Alan DeKok wrote:
> Arran Cudbard-Bell wrote:
>   
>> It was PAP...
>>
>> Running with debugging mode *off* yields much better results !
>>     
>
>   Well, yes.
>
>   
I liked the pretty text ?!
>> Still with completely vanilla install ,
>> only getting
>>
>>      [req] => 5000
>>      [parallel] => 10
>>      [total] => 50000
>>      [start] => 1178820825.99
>>      [stop] => 1178820836.79
>>      [period] => 10.791821003
>>      [req_s] => 4633
>>     
>
>   That's pretty low.  I'm a little surprised, in fact.  I was getting
> that performance almost 6 years ago on machines available then.
>
>   
Yes it's worryingly low,

This is with the G5 optimisation compiler flags passed to GCC...
Though i'm not sure it's taking them ...

should just be

env CFLAGS=-O3 -mcpu=970 -mtune=970 -mpowerpc64
./configure --prefix=/usr/local/freeradius-cvsfast

make
make install ...

But the configure script doesn't seem to take notice of the CFLAGS env 
variable , so does not pass the flags to the compiler at build time.

Have you had anyone else running FreeRADIUS on a G5 feed back about 
performance ?
>> localhost to localhost
>>     
>
>   Try from one machine to another.  Maybe the client & server are
> competing for resources on the same machine.
>   
Will try machine to machine on monday, or later today if I have time, 
though I suspect the performance will probably be worse.

top reveals that radclient is running at about 70% utilization and 
radiusd at 104%.

These boxes are not under any kind of load and have a pretty vanilla 
install of Mac OSX Server 10.4.9.
Can't figure out why it's so slow.
>   Alan DeKok.
> --
>   http://deployingradius.com       - The web site of the book
>   http://deployingradius.com/blog/ - The blog
> - 
> List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
>