From: jos poortvliet <jos@mijnkamer.nl>
To: ck@vds.kolivas.org
Cc: Con Kolivas <kernel@kolivas.org>, Ingo Molnar <mingo@elte.hu>,
Al Boldi <a1426z@gawab.com>, Mike Galbraith <efault@gmx.de>,
linux-kernel@vger.kernel.org, Nicholas Miell <nmiell@comcast.net>,
Linus Torvalds <torvalds@linux-foundation.org>,
Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [ck] Re: is RSDL an "unfair" scheduler too?
Date: Sat, 17 Mar 2007 13:23:43 +0100 [thread overview]
Message-ID: <200703171323.44517.jos@mijnkamer.nl> (raw)
In-Reply-To: <200703172302.10819.kernel@kolivas.org>
[-- Attachment #1: Type: text/plain, Size: 4560 bytes --]
Op Saturday 17 March 2007, schreef Con Kolivas:
> On Saturday 17 March 2007 22:49, Ingo Molnar wrote:
> > * Con Kolivas <kernel@kolivas.org> wrote:
> > > Despite the claims to the contrary, RSDL does not have _less_
> > > heuristics, it does not have _any_. It's purely entitlement based.
> >
> > RSDL still has heuristics very much, but this time it's hardcoded into
> > the design! Let me demonstrate this via a simple experiment.
> >
> > in the vanilla scheduler, the heuristics are ontop of a fairly basic
> > (and fast) scheduler, they are plain visible and thus 'optional'. In
> > RSDL, the heuristics are still present but more hidden and more
> > engrained into the design.
> >
> > But it's easy to demonstrate this under RSDL: consider the following two
> > scenarios, which implement precisely the same fundamental computing
> > workload (everything running on the same, default nice 0 level):
> >
> > 1) a single task runs almost all the time and sleeps about 1 msec every
> > 100 msecs.
> >
> > [ run "while N=1; do N=1; done &" under bash to create such a
> > workload. ]
> >
> > 2) tasks are in a 'ring' where each runs for 100 msec, sleeps for 1
> > msec and passes the 'token' around to the next task in the ring. (in
> > essence every task will sleep 9900 msecs before getting another run)
> >
> > [ run http://redhat.com/~mingo/scheduler-patches/ring-test.c to
> > create this workload. If the 100 tasks default is too much for you
> > then you can run "./ring-test 10" - that will show similar effects.
> > ]
> >
> > Workload #1 uses 100% of CPU time. Workload #2 uses 99% of CPU time.
> > They both do in essence the same thing.
> >
> > if RSDL had no heuristics at all then if i mixed #1 with #2, both
> > workloads would get roughly 50%/50% of the CPU, right? (as happens if i
> > mix #1 with #1 - both CPU-intense workloads get half of the CPU)
> >
> > in reality, in the 'ring workload' case, RSDL will only give about _5%_
> > of CPU time to the #1 CPU-intense task, and will give 95% of CPU time to
> > the #2 'ring' of tasks. So the distribution of timeslices is
> > significantly unfair!
> >
> > Why? Because RSDL still has heuristics, just elsewhere and more hidden:
> > in the "straightforward CPU intense task" case RSDL will 'penalize' the
> > task by depleting its quota for running nearly all the time, in the
> > "ring of tasks" case the 100 tasks will each run near their priority
> > maximum, fed by 'major epoch' events of RSDL, thus they get 'rewarded'
> > for seemingly sleeping alot and spreading things out. So RSDL has
> > fundamental unfairness built in as well - it's just different from the
> > vanilla scheduler.
>
> We're obviously disagreeing on what heuristics are so call it what you
> like.
>
> You're simply cashing in on the deep pipes that do kernel work for other
> tasks. You know very well that I dropped the TASK_NONINTERACTIVE flag from
> rsdl which checks that tasks are waiting on pipes and you're exploiting it.
> That's not the RSDL heuristics at work at all, but you're trying to make it
> look like it is the intrinsic RSDL system at work. Putting that flag back
> in is simple enough when I'm not drugged. You could have simply pointed
> that out instead of trying to make my code look responsible.
>
> For the moment I'll assume you're not simply trying to make my code look
> bad and that you thought there really was an intrinsic design problem,
> otherwise I'd really be unhappy with what was happening to me.
Well, re-reading his post, he has a point - one WOULD expect each of these 2
tasks to have an equal share of CPU, and if RSDL doesn't currently take this
pipe-thing into account, it might need some fixing. call it heuristics or not
(after all, how could one NOT say a scheduler uses heuristics of some kind?).
Anyway, relax (you know getting angry won't help you getting better) and
remember this is email - not exactly a perfet way to communicate, esp in the
emotional area. I haven't said this anywhere else, as I'm waiting for RSDL to
be a bit more mature, but I have irritations with it as well - I don't have
the long full-system stalls I had with staircase (hail RSDL!) but I do have
more frequent, shorter stalls, when one app doesn't respond for up to 10
seconds, while others just continue to work. So it's not perfect yet, and
when I have time, I'll try to find out what's wrong. BTW, nice seems to help,
but not entirely.
grtz
Jos
[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]
next prev parent reply other threads:[~2007-03-17 12:20 UTC|newest]
Thread overview: 197+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-03-04 20:35 [ANNOUNCE] RSDL completely fair starvation free interactive cpu scheduler Al Boldi
2007-03-04 21:49 ` Con Kolivas
[not found] ` <45EB45F7.3050208@simon.arlott.org.uk>
2007-03-04 22:27 ` Con Kolivas
2007-03-05 18:29 ` Simon Arlott
2007-03-05 21:36 ` Con Kolivas
2007-03-04 23:13 ` Willy Tarreau
2007-03-04 23:58 ` Con Kolivas
2007-03-05 0:10 ` [ck] " jos poortvliet
2007-03-05 1:09 ` Gene Heskett
[not found] ` <200703050834.45712.a1426z@gawab.com>
[not found] ` <20070305060732.GQ30401@nysv.org>
2007-03-05 11:59 ` [ck] " Al Boldi
2007-03-05 12:29 ` Con Kolivas
[not found] ` <200703052123.01095.a1426z@gawab.com>
2007-03-05 22:10 ` Con Kolivas
2007-03-06 8:42 ` Xavier Bestel
2007-03-06 15:15 ` Al Boldi
2007-03-11 18:11 ` Al Boldi
2007-03-11 21:52 ` Con Kolivas
2007-03-11 22:12 ` Con Kolivas
2007-03-12 4:42 ` Al Boldi
2007-03-12 4:53 ` Con Kolivas
2007-03-12 11:26 ` Al Boldi
2007-03-12 12:52 ` Con Kolivas
2007-03-12 14:14 ` Al Boldi
2007-03-12 14:58 ` [ck] " jos poortvliet
2007-03-12 16:37 ` michael chang
2007-03-12 17:41 ` Al Boldi
2007-03-12 18:05 ` Con Kolivas
2007-03-12 18:47 ` [ck] " jos poortvliet
2007-03-12 18:58 ` Antonio Vargas
2007-03-19 10:47 ` Helge Hafting
2007-03-18 1:30 ` Bill Davidsen
2007-03-18 10:50 ` [ck] " jos poortvliet
2007-03-13 15:31 ` [PATCH] [RSDL-0.30] sched: rsdl improve latencies with differential nice Con Kolivas
2007-03-13 16:03 ` [PATCH] [RSDL-0.30] sched: rsdl improve latencies with differential nice -1 Con Kolivas
2007-03-13 16:08 ` Con Kolivas
2007-03-13 20:58 ` Con Kolivas
2007-03-13 23:08 ` RSDL development plans Con Kolivas
2007-03-16 12:25 ` Con Kolivas
2007-03-16 13:40 ` RSDL v0.31 Con Kolivas
2007-03-16 15:34 ` Mike Galbraith
2007-03-16 21:13 ` Con Kolivas
2007-03-16 22:30 ` Mike Galbraith
2007-03-16 23:05 ` [ck] " Dirk Schoebel
2007-03-17 4:24 ` Nicholas Miell
2007-03-17 5:56 ` Mike Galbraith
2007-03-17 6:08 ` Mike Galbraith
2007-03-17 13:56 ` Ed Tomlinson
2007-03-18 19:37 ` Lee Revell
2007-03-18 19:55 ` Mike Galbraith
2007-03-18 22:45 ` Szonyi Calin
2007-03-19 2:27 ` David Schwartz
2007-03-19 6:21 ` Mike Galbraith
2007-03-19 6:59 ` Willy Tarreau
2007-03-17 6:26 ` Nicholas Miell
2007-03-17 7:11 ` Mike Galbraith
2007-03-17 7:25 ` William Lee Irwin III
2007-03-17 7:29 ` Nicholas Miell
2007-03-17 11:48 ` Gene Heskett
2007-03-17 7:45 ` Ingo Molnar
2007-03-17 7:44 ` David Lang
2007-03-17 8:46 ` Mike Galbraith
2007-03-17 14:09 ` [ck] " Mark Glines
2007-03-17 14:33 ` Mike Galbraith
2007-03-17 14:54 ` Mark Glines
2007-03-17 14:58 ` Mike Galbraith
2007-03-17 8:23 ` Nicholas Miell
2007-03-17 9:42 ` [patch] CFS scheduler: Completely Fair Scheduler / CONFIG_SCHED_FAIR Ingo Molnar
2007-03-17 8:41 ` RSDL v0.31 Serge Belyshev
2007-03-17 9:48 ` Con Kolivas
2007-03-17 9:58 ` Mike Galbraith
2007-03-17 10:49 ` Mike Galbraith
2007-03-17 12:05 ` Gene Heskett
2007-03-17 13:36 ` Mike Galbraith
2007-03-17 17:03 ` Gene Heskett
2007-03-17 17:37 ` Mike Galbraith
2007-03-17 18:23 ` [ck] " Kacper Wysocki
2007-03-17 18:45 ` Mike Galbraith
2007-03-17 13:58 ` michael chang
2007-03-17 20:55 ` Al Boldi
2007-03-18 6:17 ` Mike Galbraith
2007-03-18 6:47 ` Kasper Sandberg
2007-03-18 7:08 ` Mike Galbraith
2007-03-18 7:22 ` [ck] " Radoslaw Szkodzinski
2007-03-18 7:38 ` Mike Galbraith
2007-03-18 8:04 ` Mike Galbraith
2007-03-18 8:20 ` jimmy bahuleyan
2007-03-18 8:34 ` Mike Galbraith
2007-03-18 9:57 ` Kasper Sandberg
2007-03-18 13:57 ` Avuton Olrich
2007-03-19 20:47 ` Bill Davidsen
2007-03-20 10:19 ` jos poortvliet
2007-03-21 8:58 ` Kasper Sandberg
2007-03-18 15:44 ` Radoslaw Szkodzinski
2007-03-18 16:09 ` jos poortvliet
2007-03-19 16:07 ` Mark Lord
2007-03-19 16:26 ` Xavier Bestel
2007-03-19 16:36 ` Mark Lord
2007-03-19 16:43 ` Xavier Bestel
2007-03-20 3:11 ` Linus Torvalds
2007-03-20 6:11 ` Willy Tarreau
2007-03-20 8:03 ` Mike Galbraith
2007-03-21 14:57 ` Mike Galbraith
2007-03-21 16:02 ` Peter Zijlstra
2007-03-21 17:06 ` Mike Galbraith
2007-03-22 7:07 ` Mike Galbraith
2007-03-22 9:18 ` Ingo Molnar
2007-03-22 9:34 ` Mike Galbraith
2007-03-22 9:41 ` Mike Galbraith
2007-03-22 22:03 ` Con Kolivas
2007-03-22 22:50 ` Con Kolivas
2007-03-23 4:39 ` Mike Galbraith
2007-03-23 5:59 ` Con Kolivas
2007-03-23 6:11 ` Mike Galbraith
2007-03-23 12:17 ` Mike Galbraith
[not found] ` <20070321161147.54c7a727@localhost>
2007-03-21 17:07 ` Mike Galbraith
2007-03-22 4:49 ` Willy Tarreau
2007-03-22 7:14 ` Mike Galbraith
2007-03-20 9:03 ` Xavier Bestel
2007-03-20 12:31 ` Artur Skawina
2007-03-20 19:16 ` Artur Skawina
2007-03-21 7:50 ` Ingo Molnar
2007-03-21 10:43 ` David Schwartz
2007-03-28 23:37 ` Bill Davidsen
2007-03-29 7:10 ` David Schwartz
2007-03-29 7:34 ` Nick Piggin
2007-03-20 15:31 ` Linus Torvalds
2007-03-20 18:08 ` Al Boldi
2007-03-21 8:22 ` Keith Duthie
2007-03-28 23:43 ` Bill Davidsen
2007-03-20 10:26 ` [ck] " jos poortvliet
2007-03-20 13:22 ` Mark Lord
2007-03-20 15:16 ` Ray Lee
2007-03-20 15:20 ` Mark Lord
2007-03-21 8:55 ` Kasper Sandberg
2007-03-19 20:53 ` Al Boldi
2007-03-20 19:50 ` Artur Skawina
2007-03-21 4:15 ` Al Boldi
2007-03-21 17:24 ` Artur Skawina
2007-03-19 16:03 ` Mark Lord
2007-03-17 11:49 ` is RSDL an "unfair" scheduler too? Ingo Molnar
2007-03-17 12:02 ` Con Kolivas
2007-03-17 12:23 ` jos poortvliet [this message]
2007-03-17 17:31 ` [ck] " David Schwartz
2007-03-17 12:28 ` Ingo Molnar
2007-03-17 12:43 ` Con Kolivas
2007-03-17 16:34 ` Ingo Molnar
2007-03-18 3:23 ` Bill Davidsen
2007-03-18 2:13 ` Bill Davidsen
2007-03-18 3:20 ` Kasper Sandberg
2007-03-18 5:37 ` Mike Galbraith
2007-03-18 10:58 ` [ck] " jos poortvliet
2007-03-17 12:15 ` jos poortvliet
2007-03-17 20:41 ` Avi Kivity
2007-03-18 1:25 ` William Lee Irwin III
2007-03-18 1:32 ` Linus Torvalds
2007-03-18 5:24 ` Willy Tarreau
2007-03-18 5:55 ` Avi Kivity
2007-03-19 2:27 ` David Schwartz
2007-03-19 13:27 ` Radoslaw Szkodzinski
2007-03-19 18:30 ` David Lang
2007-03-19 15:25 ` Avi Kivity
2007-03-19 16:06 ` Helge Hafting
2007-03-19 16:37 ` Avi Kivity
2007-03-18 6:09 ` Bill Huey
2007-03-18 6:37 ` Mike Galbraith
2007-03-18 7:35 ` Bill Huey
2007-03-19 21:14 ` Bill Davidsen
2007-03-18 6:26 ` Mike Galbraith
2007-03-18 6:54 ` [ck] " Radoslaw Szkodzinski
2007-03-18 7:58 ` Willy Tarreau
2007-03-18 8:45 ` Avi Kivity
2007-03-18 5:00 ` Avi Kivity
2007-03-17 15:13 ` RSDL v0.31 Mark Hahn
2007-03-17 17:22 ` Stephen Clark
2007-03-19 15:06 ` Chris Friesen
2007-03-17 7:56 ` Ingo Molnar
2007-03-17 11:07 ` [ck] " jos poortvliet
2007-03-17 12:44 ` Ingo Molnar
2007-03-17 13:44 ` jos poortvliet
2007-03-17 14:04 ` [ck] " Ed Tomlinson
2007-03-17 14:32 ` Rik van Riel
2007-03-17 14:43 ` Mike Galbraith
2007-03-17 15:39 ` Ingo Molnar
2007-03-16 17:12 ` AshMilsted
2007-03-16 17:41 ` Gabriel C
2007-03-16 21:55 ` Al Boldi
2007-03-17 2:51 ` Con Kolivas
2007-03-17 4:40 ` Al Boldi
2007-03-17 4:57 ` Con Kolivas
2007-03-17 5:15 ` Gene Heskett
2007-03-17 13:50 ` Ed Tomlinson
2007-03-17 16:12 ` Al Boldi
2007-03-16 13:42 ` RSDL development plans Mike Galbraith
2007-03-16 13:59 ` Con Kolivas
2007-03-16 14:07 ` Mike Galbraith
2007-03-14 9:13 ` [PATCH] [RSDL-0.30] sched: rsdl improve latencies with differential nice Mike Galbraith
2007-03-14 9:25 ` Con Kolivas
2007-03-14 9:42 ` Mike Galbraith
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=200703171323.44517.jos@mijnkamer.nl \
--to=jos@mijnkamer.nl \
--cc=a1426z@gawab.com \
--cc=akpm@linux-foundation.org \
--cc=ck@vds.kolivas.org \
--cc=efault@gmx.de \
--cc=kernel@kolivas.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=nmiell@comcast.net \
--cc=torvalds@linux-foundation.org \
/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;
as well as URLs for NNTP newsgroup(s).