public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: David Newall <davidn@davidnewall.com>
To: Olaf Kirch <okir@suse.de>
Cc: linux-kernel@vger.kernel.org, mingo@redhat.com,
	Andreas Gruenbacher <agruen@suse.de>
Subject: Re: CFS Performance Issues
Date: Fri, 29 May 2009 01:19:06 +0930	[thread overview]
Message-ID: <4A1EB272.3050902@davidnewall.com> (raw)
In-Reply-To: <200905281502.22487.okir@suse.de>

Olaf Kirch wrote:
> As you probably know, we've been chasing a variety of performance issues
> ...
> I see this:
>
> ./slice 16
>     avg slice:  1.12 utime: 216263.187500
> ...
> Any insight you can offer here is greatly appreciated!
>   

About that: avg slice is in nsec, not msec (the old, off-by-one-million
bug), and utime, also an average, is in usec.

The first result indicates 1.12 nsec per context switch,  193 context
switches and 346% CPU utilisation.  You must have at least four CPU
cores.  Here's your table, extended* per this interpretation:

./slice 16
    avg slice:  1.12 utime: 216263.187500:  1.12 nsec/csw,  193 csw, 346 CPU%
    avg slice:  0.25 utime: 125507.687500:  0.25 nsec/csw,  502 csw, 200 CPU%
    avg slice:  0.31 utime: 125257.937500:  0.31 nsec/csw,  404 csw, 200 CPU%
    avg slice:  0.31 utime: 125507.812500:  0.31 nsec/csw,  404 csw, 200 CPU%
    avg slice:  0.12 utime: 124507.875000:  0.12 nsec/csw, 1037 csw, 199 CPU%
    avg slice:  0.38 utime: 124757.687500:  0.38 nsec/csw,  328 csw, 199 CPU%
    avg slice:  0.31 utime: 125508.000000:  0.31 nsec/csw,  404 csw, 200 CPU%
    avg slice:  0.44 utime: 125757.750000:  0.44 nsec/csw,  285 csw, 201 CPU%
    avg slice:  2.00 utime: 128258.000000:  2.00 nsec/csw,   64 csw, 205 CPU%
 ------ here I turned off new_fair_sleepers ----
    avg slice: 10.25 utime: 137008.500000: 10.25 nsec/csw,   13 csw, 219 CPU%
    avg slice:  9.31 utime: 139008.875000:  9.31 nsec/csw,   14 csw, 222 CPU%
    avg slice: 10.50 utime: 141508.687500: 10.50 nsec/csw,   13 csw, 226 CPU%
    avg slice:  9.44 utime: 139258.750000:  9.44 nsec/csw,   14 csw, 222 CPU%
    avg slice: 10.31 utime: 140008.687500: 10.31 nsec/csw,   13 csw, 224 CPU%
    avg slice:  9.19 utime: 139008.625000:  9.19 nsec/csw,   15 csw, 222 CPU%
    avg slice: 10.00 utime: 137258.625000: 10.00 nsec/csw,   13 csw, 219 CPU%
    avg slice: 10.06 utime: 135258.562500: 10.06 nsec/csw,   13 csw, 216 CPU%
    avg slice:  9.62 utime: 138758.562500:  9.62 nsec/csw,   14 csw, 222 CPU%


You don't seem to be getting good CPU utilisation.

*awk '{printf "%s: %5.2f nsec/csw, %4d csw, %3d CPU%%\n", $0, $3, $5/$3/1000, $5*16/10000}'


  reply	other threads:[~2009-05-28 15:49 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-05-28 13:02 CFS Performance Issues Olaf Kirch
2009-05-28 15:49 ` David Newall [this message]
2009-05-28 18:20   ` Olaf Kirch
2009-05-28 18:43     ` David Newall
2009-05-28 20:31 ` Peter Zijlstra
2009-05-30 11:18   ` Ingo Molnar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4A1EB272.3050902@davidnewall.com \
    --to=davidn@davidnewall.com \
    --cc=agruen@suse.de \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=okir@suse.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox