Subject: Re: High Load / bad response times
To: None <netbsd-users@netbsd.org>
From: Jochen Keil <J.Keil@gmx.de>
List: netbsd-users
Date: 01/07/2005 19:07:20
(Update)
You might find this interesting. I booted a single CPU kernel and now 
there are much less interrupts when there is disk io.

sysstat vmstat:
     5 users    Load  3.23  2.83  1.77                  Fri Jan  7 18:59:47

Proc:r  d  s  w    Csw   Trp   Sys  Int  Sof   Flt  PAGING SWAPPING
      2     5      1440 10864   759  939  557 10415  in out in out
                                                       ops
   77.0% Sy   9.6% Us   0.0% Ni   4.6% In   8.8% Id    pages
|    |    |    |    |    |    |    |    |    |    |
=======================================>>>>%%%                    forks
                                                                   fkppw
           memory totals (in kB)             1042 Interrupts       fksvm
          real   virtual    free              100 cpu0 softclock   pwait
Active 244524    244524  243640                3 cpu0 softnet     relck
All    269684    269684  375216                5 pic0 pin 7       rlkok
                                              375 pic0 pin 3       noram
Namei         Sys-cache     Proc-cache       462 pic0 pin 5       ndcpy
     Calls     hits    %     hits     %           pic0 pin 11      fltcp
        18       18  100                          pic0 pin 9    26 zfod
                                                  pic0 pin 14      cow
Disks: seeks xfers bytes %busy                97 pic0 pin 0    64 fmin
    sd0         423   26M  55.1                                 85 ftarg
    sd1                                                            itarg
    wd0         516   32M  68.7                                885 wired
    wd1                                                            pdfre
    wd2                                                            pdscn
    wd3
    cd0

Ping respone time is still relatively high at 0.2ms but i won't complain 
about that :)
Seems as if the problem lies in the SMP Kernel other than previous 
stated (maybe RAIDFframe magnified the problem so my previous non-SMP 
tests weren't successful).

Are there any options other than switching back to FreeBSD? You have to 
admit that running a SMP system without SMP support is quite senseless.

Best Regards,

Jochen Keil