Re: cpu usage

[6bone%6bone.informatik.uni-leipzig.de@localhost]

hello,

it is a single-threaded perl script.
I attached some diagnostic output, but I can't see why the perl process 
is not using the full cpu power.


Regards
Uwe


a snapshot from top (sorted by memory usage):

load averages:  1.29,  1.40,  1.42;               up 11+02:15:23       14:23:43
123 processes: 121 sleeping, 2 on CPU
CPU0 states:  1.2% user,  0.0% nice,  1.2% system,  4.0% interrupt, 93.6% idle
CPU1 states:  1.6% user,  0.0% nice,  0.4% system,  0.0% interrupt, 98.0% idle
CPU2 states:  0.4% user,  0.0% nice,  3.2% system,  0.0% interrupt, 96.4% idle
CPU3 states:  0.0% user,  0.0% nice,  2.2% system,  0.0% interrupt, 97.8% idle
Memory: 9439M Act, 4608M Inact, 9356K Wired, 23M Exec, 7024M File, 12M Free
Swap: 71G Total, 16G Used, 55G Free

  PID USERNAME PRI NICE   SIZE   RES STATE      TIME   WCPU    CPU COMMAND
11146 steuck    93  -15    16G 6955M biowai/2  17:39  0.00%  0.00% perl
25321 mirror    85    0    50M 1940K select/1   2:13  0.00%  0.00% rsync
  533 root      85    0    47M   12M select/2   1:48  0.00%  0.00% httpd
15856 www       85    0    47M   11M select/2   0:00  0.00%  0.00% httpd
...


iostat -D 1
              ld0               ld1               ld2               ld3
    KB   xfr time     KB   xfr time     KB   xfr time     KB   xfr time
   956    29 0.02     50     3 0.00    152    25 0.06    153    23 0.05
  3032    79 0.06      0     0 0.00    297    88 0.60    260    77 0.49
  2788    75 0.09      0     0 0.00    329    89 0.55    318    87 0.52
  2962    75 0.05      8     1 0.00    263    74 0.49    327    91 0.55
  2919    74 0.06      0     0 0.00    362   100 0.63    330    90 0.42
  2630    66 0.10      0     0 0.00    369   101 0.71    293    84 0.41
  2725    71 0.11      0     0 0.00    294    78 0.59    260    71 0.50
  3331    84 0.21      0     0 0.00    290    81 0.58    336    94 0.55
....



sysstat vm (snapshot):

   11 users    Load  1.36  1.48  1.45                  Mon Oct  5 14:29:33

Proc:r  d  s  w     Csw    Trp    Sys   Int   Sof    Flt      PAGING   SWAPPING
          95       5859   3214   5535  1506  1773   3209      in  out   in  out
                                                        ops  163   22
   2.2% Sy   0.4% Us   0.0% Ni   0.6% In  96.8% Id    pages       350
|    |    |    |    |    |    |    |    |    |    |
=%                                                                      4 forks
                                                                          fkppw
           memory totals (in kB)            2565 Interrupts               fksvm
          real  virtual     free            1059 global TLB IPI           pwait
Active 9665188 26715992     7908               7 ioapic1 pin 14       163 relck
All   16241424 33292228 57259540            1199 ioapic0 pin 16       163 rlkok
                                             300 ioapic0 pin 18           noram
Namei         Sys-cache     Proc-cache           ioapic0 pin 21       131 ndcpy
    Calls     hits    %     hits     %           ioapic0 pin 20        32 fltcp
     3329     3307   99       14     0           ioapic0 pin 14       648 zfod
                                                 ioapic0 pin 4        436 cow
Disks:   ld0   ld1   ld2   ld3   cd0                                 1024 fmin
 seeks                                                               1365 ftarg
 xfers    62     1   136   112                                            itarg
 bytes 2191K   48K 1041K 1012K                                       2394 wired
 %busy   7.0   0.6  59.2  51.2                                        821 pdfre



On Mon, 5 Oct 2009, Matthias Scheler wrote:

> Date: Mon, 5 Oct 2009 13:21:14 +0100
> From: Matthias Scheler <tron%zhadum.org.uk@localhost>
> To: 6bone%6bone.informatik.uni-leipzig.de@localhost
> Cc: netbsd-users%netbsd.org@localhost
> Subject: Re: cpu usage
>
> On Mon, Oct 05, 2009 at 01:55:54PM +0200, 
> 6bone%6bone.informatik.uni-leipzig.de@localhost wrote:
> > WAPBL is compiled into the kernel but not used.
>
> It can't be the cause of the problem in that case. WAPBL can degrade
> performance in some cases when the VM sub-systems needs to free
> a lot pages.
>
> > Can it be a problem with the size of the process?
>
> Maybe, but I don't know for sure. Is it a multi-threaded Perl script?
> Perls threading has its locking issues.
>
> Did you try running "systat vm" to see what is going on?
>
>         Kind regards
>
> --
> Matthias Scheler                                  http://zhadum.org.uk/