lkml.org 
[lkml]   [1999]   [Nov]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: stressing linux
I'm quite currious. The time you give of user vs. system is sampled
on the system clock (this is how it's done one linux). The problem
is that this data is highly inaccurate. Events in the kernel happen
every couple of microseconds. Sampling on the system clock, which
happens every 10 ms on an intel system, leaves 10000 microseconds
unaccounted for! Have you tried the Linux Trace Toolkit? This tool
will provide you with the exact times, down to the microsecond of
usage for any process and exactly how much time this process lost
waiting for I/O. This is information that is currently unavailable
by any other means.

LTT can be found at :

http://www.info.polymtl.ca/~karym/trace

Regards.

Raul Miller wrote:
>
> $ time perl -le 'for (1..80000){open FILE, ">$_" or die $!}'
>
> real 26m51.912s
> user 0m3.250s
> sys 26m35.630s
> $
>
> While this was running, X performance sucked.
>
> There's a variety of ways of classifying this problem. Among other
> things we're classifying a compute bound process as i/o bound.
>
> Could someone put this into their suite of stress tests?
>
> * * * * *
>
> By the way, for comparison purposes:
>
> $ time rm -rf t
>
> real 2m9.219s
> user 0m0.340s
> sys 2m7.510s
> $
>
> ("t" was the directory where I created all those files).
>
> And,
>
> $ time perl -le 'use File::Basename; for (1..80000){s-\B-/-g; s/$/f/; $d= dirname $_; -d $d or mkdir $d, 0777; open FILE, ">$_" or die $!}'
>
> real 1m29.630s
> user 0m10.940s
> sys 0m5.100s
> $ cd ..
> $ time rm -rf t
>
> real 0m2.816s
> user 0m0.450s
> sys 0m2.370s
> $
>
> [Plus maybe half a minute of disk activity after a short delay, as all
> those sub-directory entries were cleaned up.]
>
> --
> Raul
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/

--
===================================================
Karim Yaghmour
karym@info.polymtl.ca
Operating System Consultant
(Linux kernel, real-time and distributed systems)
===================================================

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:55    [W:0.113 / U:0.080 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site