From: peterz@infradead•org (Peter Zijlstra)
To: linux-arm-kernel@lists•infradead.org
Subject: [PATCH v2 08/11] sched: get CPU's activity statistic
Date: Wed, 4 Jun 2014 12:17:24 +0200 [thread overview]
Message-ID: <20140604101724.GD11096@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <CAKfTPtBCG1Jq2b+fyoGuA=7yG_yUbsmaD+j-ZXXsPNGp5giKpg@mail.gmail.com>
On Wed, Jun 04, 2014 at 11:32:10AM +0200, Vincent Guittot wrote:
> On 4 June 2014 10:08, Peter Zijlstra <peterz@infradead•org> wrote:
> > On Wed, Jun 04, 2014 at 09:47:26AM +0200, Vincent Guittot wrote:
> >> On 3 June 2014 17:50, Peter Zijlstra <peterz@infradead•org> wrote:
> >> > On Wed, May 28, 2014 at 04:47:03PM +0100, Morten Rasmussen wrote:
> >> >> Since we may do periodic load-balance every 10 ms or so, we will perform
> >> >> a number of load-balances where runnable_avg_sum will mostly be
> >> >> reflecting the state of the world before a change (new task queued or
> >> >> moved a task to a different cpu). If you had have two tasks continuously
> >> >> on one cpu and your other cpu is idle, and you move one of the tasks to
> >> >> the other cpu, runnable_avg_sum will remain unchanged, 47742, on the
> >> >> first cpu while it starts from 0 on the other one. 10 ms later it will
> >> >> have increased a bit, 32 ms later it will be 47742/2, and 345 ms later
> >> >> it reaches 47742. In the mean time the cpu doesn't appear fully utilized
> >> >> and we might decide to put more tasks on it because we don't know if
> >> >> runnable_avg_sum represents a partially utilized cpu (for example a 50%
> >> >> task) or if it will continue to rise and eventually get to 47742.
> >> >
> >> > Ah, no, since we track per task, and update the per-cpu ones when we
> >> > migrate tasks, the per-cpu values should be instantly updated.
> >> >
> >> > If we were to increase per task storage, we might as well also track
> >> > running_avg not only runnable_avg.
> >>
> >> I agree that the removed running_avg should give more useful
> >> information about the the load of a CPU.
> >>
> >> The main issue with running_avg is that it's disturbed by other tasks
> >> (as point out previously). As a typical example, if we have 2 tasks
> >> with a load of 25% on 1 CPU, the unweighted runnable_load_avg will be
> >> in the range of [100% - 50%] depending of the parallelism of the
> >> runtime of the tasks whereas the reality is 50% and the use of
> >> running_avg will return this value
> >
> > I'm not sure I see how 100% is possible, but yes I agree that runnable
> > can indeed be inflated due to this queueing effect.
Let me explain the 75%, take any one of the above scenarios. Lets call
the two tasks A and B, and let for a moment assume A always wins and
runs first, and then B.
So A will be runnable for 25%, B otoh will be runnable the entire time A
is actually running plus its own running time, giving 50%. Together that
makes 75%.
If you release the assumption that A runs first, but instead assume they
equally win the first execution, you get them averaging at 37.5% each,
which combined will still give 75%.
> In fact, it can be even worse than that because i forgot to take into
> account the geometric series effect which implies that it depends of
> the runtime (idletime) of the task
>
> Take 3 examples:
>
> 2 tasks that need to run 10ms simultaneously each 40ms. If they share
> the same CPU, they will be on the runqueue 20ms (in fact a bit less
> for one of them), Their load (runnable_avg_sum/runnable_avg_period)
> will be 33% each so the unweighted runnable_load_avg of the CPU will
> be 66%
>
> 2 tasks that need to run 25ms simultaneously each 100ms. If they share
> the same CPU, they will be on the runqueue 50ms (in fact a bit less
> for one of them), Their load (runnable_avg_sum/runnable_avg_period)
> will be 74% each so the unweighted runnable_load_avg of the CPU will
> be 148%
>
> 2 tasks that need to run 50ms simultaneously each 200ms. If they
> share the same CPU, they will be on the runqueue 100ms (in fact a bit
> less for one of them), Their load
> (runnable_avg_sum/runnable_avg_period) will be 89% each so the
> unweighted runnable_load_avg of the CPU will be 180%
And this is because the running time is 'large' compared to the decay
and we get hit by the weight of the recent state? Yes, I can see that,
the avg will fluctuate due to the nature of this thing.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20140604/482868c4/attachment.sig>
next prev parent reply other threads:[~2014-06-04 10:17 UTC|newest]
Thread overview: 108+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-05-23 15:52 [PATCH v2 00/11] sched: consolidation of cpu_power Vincent Guittot
2014-05-23 15:52 ` [PATCH v2 01/11] sched: fix imbalance flag reset Vincent Guittot
2014-05-25 10:33 ` Preeti U Murthy
2014-05-26 7:49 ` Vincent Guittot
2014-05-26 9:16 ` Preeti U Murthy
2014-05-26 10:14 ` Vincent Guittot
2014-05-23 15:52 ` [PATCH v2 02/11] sched: remove a wake_affine condition Vincent Guittot
2014-05-27 12:48 ` Peter Zijlstra
2014-05-27 15:19 ` Vincent Guittot
2014-05-27 15:39 ` Peter Zijlstra
2014-05-27 16:14 ` Vincent Guittot
2014-05-28 6:49 ` Vincent Guittot
2014-05-28 15:09 ` Dietmar Eggemann
2014-05-28 15:25 ` Vincent Guittot
2014-05-27 13:43 ` Peter Zijlstra
2014-05-27 13:45 ` Peter Zijlstra
2014-05-27 15:20 ` Vincent Guittot
2014-05-23 15:52 ` [PATCH v2 03/11] sched: fix avg_load computation Vincent Guittot
2014-05-27 13:48 ` Peter Zijlstra
2014-05-23 15:52 ` [PATCH v2 04/11] sched: Allow all archs to set the power_orig Vincent Guittot
2014-05-30 14:04 ` Dietmar Eggemann
2014-05-30 14:46 ` Peter Zijlstra
2014-05-30 20:50 ` Vincent Guittot
2014-06-04 9:42 ` Dietmar Eggemann
2014-06-04 11:15 ` Vincent Guittot
2014-06-05 8:59 ` Dietmar Eggemann
2014-06-16 9:01 ` Vincent Guittot
2014-06-03 13:22 ` Morten Rasmussen
2014-06-03 14:02 ` Vincent Guittot
2014-06-04 11:17 ` Morten Rasmussen
2014-06-06 7:01 ` Vincent Guittot
2014-05-23 15:52 ` [PATCH v2 05/11] ARM: topology: use new cpu_power interface Vincent Guittot
2014-05-25 13:22 ` Preeti U Murthy
2014-05-26 8:25 ` Vincent Guittot
2014-05-26 9:19 ` Preeti U Murthy
2014-05-23 15:53 ` [PATCH v2 06/11] sched: add per rq cpu_power_orig Vincent Guittot
2014-05-23 15:53 ` [PATCH v2 07/11] Revert "sched: Put rq's sched_avg under CONFIG_FAIR_GROUP_SCHED" Vincent Guittot
2014-05-23 15:53 ` [PATCH v2 08/11] sched: get CPU's activity statistic Vincent Guittot
2014-05-27 17:32 ` Peter Zijlstra
2014-05-28 7:01 ` Vincent Guittot
2014-05-28 12:10 ` Morten Rasmussen
2014-05-28 13:15 ` Vincent Guittot
2014-05-28 15:47 ` Morten Rasmussen
2014-05-28 16:39 ` Vincent Guittot
2014-06-03 12:03 ` Morten Rasmussen
2014-06-03 15:59 ` Peter Zijlstra
2014-06-03 17:41 ` Morten Rasmussen
2014-06-03 15:50 ` Peter Zijlstra
2014-06-03 17:20 ` Morten Rasmussen
2014-06-04 7:47 ` Vincent Guittot
2014-06-04 8:08 ` Peter Zijlstra
2014-06-04 8:55 ` Morten Rasmussen
2014-06-04 9:23 ` Peter Zijlstra
2014-06-04 9:35 ` Vincent Guittot
2014-06-04 10:25 ` Morten Rasmussen
2014-06-04 9:44 ` Morten Rasmussen
2014-06-04 9:32 ` Vincent Guittot
2014-06-04 10:00 ` Morten Rasmussen
2014-06-04 10:17 ` Peter Zijlstra [this message]
2014-06-04 10:36 ` Morten Rasmussen
2014-06-04 10:55 ` Peter Zijlstra
2014-06-04 11:07 ` Vincent Guittot
2014-06-04 11:23 ` Morten Rasmussen
2014-06-04 11:52 ` Vincent Guittot
2014-06-04 13:09 ` Morten Rasmussen
2014-06-04 13:23 ` Morten Rasmussen
2014-05-28 15:17 ` Peter Zijlstra
2014-06-03 15:40 ` Peter Zijlstra
2014-06-03 17:16 ` Morten Rasmussen
2014-06-03 17:37 ` Peter Zijlstra
2014-06-03 17:39 ` Peter Zijlstra
2014-06-03 23:11 ` Yuyang Du
2014-05-30 9:50 ` Dietmar Eggemann
2014-05-30 19:20 ` Vincent Guittot
2014-06-01 11:33 ` Dietmar Eggemann
2014-06-02 14:07 ` Vincent Guittot
2014-05-23 15:53 ` [PATCH v2 09/11] sched: test the cpu's capacity in wake affine Vincent Guittot
2014-05-28 10:58 ` Peter Zijlstra
2014-05-28 11:15 ` Vincent Guittot
2014-11-24 0:34 ` Wanpeng Li
2014-11-24 13:23 ` Vincent Guittot
2014-05-23 15:53 ` [PATCH v2 10/11] sched: move cfs task on a CPU with higher capacity Vincent Guittot
2014-05-29 9:50 ` Peter Zijlstra
2014-05-29 19:37 ` Vincent Guittot
2014-05-30 6:29 ` Peter Zijlstra
2014-05-30 20:05 ` Vincent Guittot
2014-06-02 17:06 ` Vincent Guittot
2014-06-03 11:15 ` Peter Zijlstra
2014-06-03 12:31 ` Vincent Guittot
2014-05-29 14:04 ` Peter Zijlstra
2014-05-29 19:44 ` Vincent Guittot
2014-05-30 13:26 ` Dietmar Eggemann
2014-05-30 19:24 ` Vincent Guittot
2014-05-30 19:45 ` Nicolas Pitre
2014-05-30 20:07 ` Vincent Guittot
2014-05-23 15:53 ` [PATCH v2 11/11] sched: replace capacity by activity Vincent Guittot
2014-05-29 13:55 ` Peter Zijlstra
2014-05-29 19:51 ` Vincent Guittot
2014-06-02 6:21 ` Preeti U Murthy
2014-06-03 9:50 ` Vincent Guittot
2014-05-29 14:02 ` Peter Zijlstra
2014-05-29 19:56 ` Vincent Guittot
2014-05-30 6:34 ` Peter Zijlstra
2014-05-30 19:13 ` Vincent Guittot
2014-05-26 9:44 ` [PATCH v2 00/11] sched: consolidation of cpu_power Preeti U Murthy
2014-05-26 10:04 ` Vincent Guittot
2014-05-26 15:54 ` Vincent Guittot
2014-05-27 5:47 ` Preeti U Murthy
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=20140604101724.GD11096@twins.programming.kicks-ass.net \
--to=peterz@infradead$(echo .)org \
--cc=linux-arm-kernel@lists$(echo .)infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
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