From: Mike Galbraith <efault@gmx.de>
To: Jens Axboe <jens.axboe@oracle.com>
Cc: Ingo Molnar <mingo@elte.hu>,
Linus Torvalds <torvalds@linux-foundation.org>,
Vivek Goyal <vgoyal@redhat.com>,
Ulrich Lukas <stellplatz-nr.13a@datenparkplatz.de>,
linux-kernel@vger.kernel.org,
containers@lists.linux-foundation.org, dm-devel@redhat.com,
nauman@google.com, dpshah@google.com, lizf@cn.fujitsu.com,
mikew@google.com, fchecconi@gmail.com, paolo.valente@unimore.it,
ryov@valinux.co.jp, fernando@oss.ntt.co.jp, jmoyer@redhat.com,
dhaval@linux.vnet.ibm.com, balbir@linux.vnet.ibm.com,
righi.andrea@gmail.com, m-ikeda@ds.jp.nec.com, agk@redhat.com,
akpm@linux-foundation.org, peterz@infradead.org,
jmarchan@redhat.com, riel@redhat.com
Subject: tweaking IO latency [was Re: IO scheduler based IO controller V10]
Date: Sat, 03 Oct 2009 08:31:06 +0200 [thread overview]
Message-ID: <1254551466.7716.4.camel@marge.simson.net> (raw)
In-Reply-To: <1254549378.8299.21.camel@marge.simson.net>
P.S. now may be a good time to finally exit thread (and maybe trim cc?)
On Sat, 2009-10-03 at 07:56 +0200, Mike Galbraith wrote:
> On Sat, 2009-10-03 at 07:49 +0200, Mike Galbraith wrote:
> > On Fri, 2009-10-02 at 20:19 +0200, Jens Axboe wrote:
> >
> > > If you could do a cleaned up version of your overload patch based on
> > > this:
> > >
> > > http://git.kernel.dk/?p=linux-2.6-block.git;a=commit;h=1d2235152dc745c6d94bedb550fea84cffdbf768
> > >
> > > then lets take it from there.
>
> Note to self: build the darn thing after last minute changes.
>
> Block: Delay overloading of CFQ queues to improve read latency.
>
> Introduce a delay maximum dispatch timestamp, and stamp it when:
> 1. we encounter a known seeky or possibly new sync IO queue.
> 2. the current queue may go idle and we're draining async IO.
> 3. we have sync IO in flight and are servicing an async queue.
> 4 we are not the sole user of disk.
> Disallow exceeding quantum if any of these events have occurred recently.
>
> Protect this behavioral change with a "desktop_dispatch" knob and default
> it to "on".. providing an easy means of regression verification prior to
> hate-mail dispatch :) to CC list.
>
> Signed-off-by: Mike Galbraith <efault@gmx.de>
> Cc: Jens Axboe <jens.axboe@oracle.com>
> Cc: Linus Torvalds <torvalds@linux-foundation.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> ... others who let somewhat hacky tweak slip by
>
> ---
> block/cfq-iosched.c | 45 +++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 41 insertions(+), 4 deletions(-)
>
> Index: linux-2.6/block/cfq-iosched.c
> ===================================================================
> --- linux-2.6.orig/block/cfq-iosched.c
> +++ linux-2.6/block/cfq-iosched.c
> @@ -174,6 +174,9 @@ struct cfq_data {
> unsigned int cfq_slice_async_rq;
> unsigned int cfq_slice_idle;
> unsigned int cfq_desktop;
> + unsigned int cfq_desktop_dispatch;
> +
> + unsigned long desktop_dispatch_ts;
>
> struct list_head cic_list;
>
> @@ -1283,6 +1286,7 @@ static int cfq_dispatch_requests(struct
> struct cfq_data *cfqd = q->elevator->elevator_data;
> struct cfq_queue *cfqq;
> unsigned int max_dispatch;
> + unsigned long delay;
>
> if (!cfqd->busy_queues)
> return 0;
> @@ -1297,19 +1301,26 @@ static int cfq_dispatch_requests(struct
> /*
> * Drain async requests before we start sync IO
> */
> - if (cfq_cfqq_idle_window(cfqq) && cfqd->rq_in_driver[BLK_RW_ASYNC])
> + if (cfq_cfqq_idle_window(cfqq) && cfqd->rq_in_driver[BLK_RW_ASYNC]) {
> + cfqd->desktop_dispatch_ts = jiffies;
> return 0;
> + }
>
> /*
> * If this is an async queue and we have sync IO in flight, let it wait
> */
> - if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq))
> + if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq)) {
> + cfqd->desktop_dispatch_ts = jiffies;
> return 0;
> + }
>
> max_dispatch = cfqd->cfq_quantum;
> if (cfq_class_idle(cfqq))
> max_dispatch = 1;
>
> + if (cfqd->busy_queues > 1)
> + cfqd->desktop_dispatch_ts = jiffies;
> +
> /*
> * Does this cfqq already have too much IO in flight?
> */
> @@ -1327,6 +1338,16 @@ static int cfq_dispatch_requests(struct
> return 0;
>
> /*
> + * Don't start overloading until we've been alone for a bit.
> + */
> + if (cfqd->cfq_desktop_dispatch) {
> + delay = cfqd->desktop_dispatch_ts + cfq_slice_sync;
> +
> + if (time_before(jiffies, max_delay))
> + return 0;
> + }
> +
> + /*
> * we are the only queue, allow up to 4 times of 'quantum'
> */
> if (cfqq->dispatched >= 4 * max_dispatch)
> @@ -1942,7 +1963,7 @@ static void
> cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
> struct cfq_io_context *cic)
> {
> - int old_idle, enable_idle;
> + int old_idle, enable_idle, seeky = 0;
>
> /*
> * Don't idle for async or idle io prio class
> @@ -1950,10 +1971,20 @@ cfq_update_idle_window(struct cfq_data *
> if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
> return;
>
> + if (cfqd->hw_tag) {
> + if (CIC_SEEKY(cic))
> + seeky = 1;
> + /*
> + * If seeky or incalculable seekiness, delay overloading.
> + */
> + if (seeky || !sample_valid(cic->seek_samples))
> + cfqd->desktop_dispatch_ts = jiffies;
> + }
> +
> enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
>
> if (!atomic_read(&cic->ioc->nr_tasks) || !cfqd->cfq_slice_idle ||
> - (!cfqd->cfq_desktop && cfqd->hw_tag && CIC_SEEKY(cic)))
> + (!cfqd->cfq_desktop && seeky))
> enable_idle = 0;
> else if (sample_valid(cic->ttime_samples)) {
> if (cic->ttime_mean > cfqd->cfq_slice_idle)
> @@ -2483,6 +2514,9 @@ static void *cfq_init_queue(struct reque
> cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
> cfqd->cfq_slice_idle = cfq_slice_idle;
> cfqd->cfq_desktop = 1;
> + cfqd->cfq_desktop_dispatch = 1;
> +
> + cfqd->desktop_dispatch_ts = INITIAL_JIFFIES;
> cfqd->hw_tag = 1;
>
> return cfqd;
> @@ -2553,6 +2587,7 @@ SHOW_FUNCTION(cfq_slice_sync_show, cfqd-
> SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
> SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
> SHOW_FUNCTION(cfq_desktop_show, cfqd->cfq_desktop, 0);
> +SHOW_FUNCTION(cfq_desktop_dispatch_show, cfqd->cfq_desktop_dispatch, 0);
> #undef SHOW_FUNCTION
>
> #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \
> @@ -2585,6 +2620,7 @@ STORE_FUNCTION(cfq_slice_async_store, &c
> STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
> UINT_MAX, 0);
> STORE_FUNCTION(cfq_desktop_store, &cfqd->cfq_desktop, 0, 1, 0);
> +STORE_FUNCTION(cfq_desktop_dispatch_store, &cfqd->cfq_desktop_dispatch, 0, 1, 0);
> #undef STORE_FUNCTION
>
> #define CFQ_ATTR(name) \
> @@ -2601,6 +2637,7 @@ static struct elv_fs_entry cfq_attrs[] =
> CFQ_ATTR(slice_async_rq),
> CFQ_ATTR(slice_idle),
> CFQ_ATTR(desktop),
> + CFQ_ATTR(desktop_dispatch),
> __ATTR_NULL
> };
>
>
next prev parent reply other threads:[~2009-10-03 6:31 UTC|newest]
Thread overview: 177+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-09-24 19:25 IO scheduler based IO controller V10 Vivek Goyal
2009-09-24 19:25 ` [PATCH 01/28] io-controller: Documentation Vivek Goyal
2009-09-24 19:25 ` [PATCH 02/28] io-controller: Core of the elevator fair queuing Vivek Goyal
2009-09-24 19:25 ` [PATCH 03/28] io-controller: Keep a cache of recently expired queues Vivek Goyal
2009-09-24 19:25 ` [PATCH 04/28] io-controller: Common flat fair queuing code in elevaotor layer Vivek Goyal
2009-09-24 19:25 ` [PATCH 05/28] io-controller: Modify cfq to make use of flat elevator fair queuing Vivek Goyal
2009-09-24 19:25 ` [PATCH 06/28] io-controller: Core scheduler changes to support hierarhical scheduling Vivek Goyal
2009-09-24 19:25 ` [PATCH 07/28] io-controller: cgroup related changes for hierarchical group support Vivek Goyal
2009-09-24 19:25 ` [PATCH 08/28] io-controller: Common hierarchical fair queuing code in elevaotor layer Vivek Goyal
2009-09-24 19:25 ` [PATCH 09/28] io-controller: cfq changes to use " Vivek Goyal
2009-09-24 19:25 ` [PATCH 10/28] io-controller: Export disk time used and nr sectors dipatched through cgroups Vivek Goyal
2009-09-24 19:25 ` [PATCH 11/28] io-controller: Debug hierarchical IO scheduling Vivek Goyal
2009-09-24 19:25 ` [PATCH 12/28] io-controller: Introduce group idling Vivek Goyal
2009-09-24 19:25 ` [PATCH 13/28] io-controller: Implement wait busy for io queues Vivek Goyal
2009-09-24 19:25 ` [PATCH 14/28] io-controller: Keep track of late preemptions Vivek Goyal
2009-09-24 19:25 ` [PATCH 15/28] io-controller: Allow CFQ specific extra preemptions Vivek Goyal
2009-09-25 6:24 ` Gui Jianfeng
2009-09-24 19:25 ` [PATCH 16/28] io-controller: Wait for requests to complete from last queue before new queue is scheduled Vivek Goyal
2009-09-24 19:25 ` [PATCH 17/28] io-controller: Separate out queue and data Vivek Goyal
2009-09-24 19:25 ` [PATCH 18/28] io-conroller: Prepare elevator layer for single queue schedulers Vivek Goyal
2009-09-24 19:25 ` [PATCH 19/28] io-controller: Avoid expiring ioq for single ioq scheduler if only root group Vivek Goyal
2009-09-24 19:25 ` [PATCH 20/28] io-controller: noop changes for hierarchical fair queuing Vivek Goyal
2009-09-24 19:25 ` [PATCH 21/28] io-controller: deadline " Vivek Goyal
2009-09-24 19:25 ` [PATCH 22/28] io-controller: anticipatory " Vivek Goyal
2009-09-24 19:25 ` [PATCH 23/28] io-controller: blkio_cgroup patches from Ryo to track async bios Vivek Goyal
2009-09-24 19:25 ` [PATCH 24/28] io-controller: map async requests to appropriate cgroup Vivek Goyal
2009-09-24 19:25 ` [PATCH 25/28] io-controller: Per cgroup request descriptor support Vivek Goyal
2009-09-24 19:25 ` [PATCH 26/28] io-controller: Per io group bdi congestion interface Vivek Goyal
2009-09-24 19:25 ` [PATCH 27/28] io-controller: Support per cgroup per device weights and io class Vivek Goyal
2009-09-24 19:25 ` [PATCH 28/28] io-controller: debug elevator fair queuing support Vivek Goyal
2009-09-24 21:33 ` IO scheduler based IO controller V10 Andrew Morton
2009-09-25 1:09 ` KAMEZAWA Hiroyuki
2009-09-25 1:18 ` KAMEZAWA Hiroyuki
2009-09-25 5:29 ` Balbir Singh
2009-09-25 7:09 ` Ryo Tsuruta
2009-09-25 4:14 ` Vivek Goyal
2009-09-25 5:04 ` Vivek Goyal
2009-09-25 9:07 ` Ryo Tsuruta
2009-09-25 14:33 ` Vivek Goyal
2009-09-28 7:30 ` Ryo Tsuruta
2009-09-25 15:04 ` Rik van Riel
2009-09-28 7:38 ` Ryo Tsuruta
2009-10-08 4:42 ` More performance numbers (Was: Re: IO scheduler based IO controller V10) Vivek Goyal
2009-10-08 8:34 ` Andrea Righi
2009-10-10 19:53 ` Performance numbers with IO throttling patches " Vivek Goyal
2009-10-10 22:27 ` Andrea Righi
2009-10-11 12:32 ` Vivek Goyal
2009-10-12 21:11 ` Vivek Goyal
2009-10-17 15:18 ` Andrea Righi
2009-09-25 2:20 ` IO scheduler based IO controller V10 Ulrich Lukas
2009-09-25 20:26 ` Vivek Goyal
2009-09-26 14:51 ` Mike Galbraith
2009-09-27 6:55 ` Mike Galbraith
2009-09-27 16:42 ` Jens Axboe
2009-09-27 18:15 ` Mike Galbraith
2009-09-28 4:04 ` Mike Galbraith
2009-09-28 5:55 ` Mike Galbraith
2009-09-28 17:48 ` Vivek Goyal
2009-09-28 18:24 ` Mike Galbraith
2009-09-30 19:58 ` Mike Galbraith
2009-09-30 20:05 ` Mike Galbraith
2009-09-30 20:24 ` Vivek Goyal
2009-10-01 7:33 ` Mike Galbraith
2009-10-01 18:58 ` Jens Axboe
2009-10-02 6:23 ` Mike Galbraith
2009-10-02 8:04 ` Jens Axboe
2009-10-02 8:53 ` Mike Galbraith
2009-10-02 9:00 ` Mike Galbraith
2009-10-02 9:55 ` Jens Axboe
2009-10-02 12:22 ` Mike Galbraith
2009-10-02 9:24 ` Ingo Molnar
2009-10-02 9:28 ` Jens Axboe
2009-10-02 14:24 ` Linus Torvalds
2009-10-02 14:45 ` Mike Galbraith
2009-10-02 14:57 ` Jens Axboe
2009-10-02 14:56 ` Jens Axboe
2009-10-02 15:14 ` Linus Torvalds
2009-10-02 16:01 ` jim owens
2009-10-02 17:11 ` Jens Axboe
2009-10-02 17:20 ` Ingo Molnar
2009-10-02 17:25 ` Jens Axboe
2009-10-02 17:28 ` Ingo Molnar
2009-10-02 17:37 ` Jens Axboe
2009-10-02 17:56 ` Ingo Molnar
2009-10-02 18:04 ` Jens Axboe
2009-10-02 18:22 ` Mike Galbraith
2009-10-02 18:26 ` Jens Axboe
2009-10-02 18:33 ` Mike Galbraith
2009-10-02 18:36 ` Theodore Tso
2009-10-02 18:45 ` Jens Axboe
2009-10-02 19:01 ` Ingo Molnar
2009-10-02 19:09 ` Jens Axboe
2009-10-02 18:13 ` Mike Galbraith
2009-10-02 18:19 ` Jens Axboe
2009-10-02 18:57 ` Mike Galbraith
2009-10-02 20:47 ` Mike Galbraith
2009-10-03 5:48 ` Mike Galbraith
2009-10-03 5:56 ` Mike Galbraith
2009-10-03 6:31 ` Mike Galbraith [this message]
2009-10-03 7:24 ` Jens Axboe
2009-10-03 9:00 ` Mike Galbraith
2009-10-03 9:12 ` Corrado Zoccolo
2009-10-03 13:18 ` Jens Axboe
2009-10-03 13:17 ` Jens Axboe
2009-10-03 11:29 ` Vivek Goyal
2009-10-03 12:40 ` Do not overload dispatch queue (Was: Re: IO scheduler based IO controller V10) Vivek Goyal
2009-10-03 13:21 ` Jens Axboe
2009-10-03 13:56 ` Vivek Goyal
2009-10-03 14:02 ` Mike Galbraith
2009-10-03 14:28 ` Jens Axboe
2009-10-03 14:33 ` Mike Galbraith
2009-10-03 14:51 ` Mike Galbraith
2009-10-03 15:14 ` Jens Axboe
2009-10-03 15:57 ` Mike Galbraith
2009-10-03 17:35 ` Jens Axboe
2009-10-03 17:45 ` Linus Torvalds
2009-10-03 17:51 ` Jens Axboe
2009-10-03 19:07 ` Mike Galbraith
2009-10-03 19:11 ` Mike Galbraith
2009-10-03 19:23 ` Jens Axboe
2009-10-03 19:49 ` Mike Galbraith
2009-10-04 10:50 ` Mike Galbraith
2009-10-04 11:33 ` Mike Galbraith
2009-10-04 17:39 ` Jens Axboe
2009-10-04 18:23 ` Mike Galbraith
2009-10-04 18:38 ` Jens Axboe
2009-10-04 19:47 ` Mike Galbraith
2009-10-04 20:17 ` Jens Axboe
2009-10-04 22:15 ` Mike Galbraith
2009-10-03 13:57 ` Mike Galbraith
2009-10-03 7:20 ` IO scheduler based IO controller V10 Ingo Molnar
2009-10-03 7:25 ` Jens Axboe
2009-10-03 8:53 ` Mike Galbraith
2009-10-03 9:01 ` Corrado Zoccolo
2009-10-02 16:33 ` Ray Lee
2009-10-02 17:13 ` Jens Axboe
2009-10-02 16:22 ` Ingo Molnar
2009-10-02 9:36 ` Mike Galbraith
2009-10-02 16:37 ` Ingo Molnar
2009-10-02 18:08 ` Jens Axboe
2009-10-02 18:29 ` Mike Galbraith
2009-10-02 18:36 ` Jens Axboe
2009-09-27 17:00 ` Corrado Zoccolo
2009-09-28 14:56 ` Vivek Goyal
2009-09-28 15:35 ` Corrado Zoccolo
2009-09-28 17:14 ` Vivek Goyal
2009-09-29 7:10 ` Corrado Zoccolo
2009-09-28 17:51 ` Mike Galbraith
2009-09-28 18:18 ` Vivek Goyal
2009-09-28 18:53 ` Mike Galbraith
2009-09-29 7:14 ` Corrado Zoccolo
2009-09-29 5:55 ` Mike Galbraith
2009-09-29 0:37 ` Nauman Rafique
2009-09-29 3:22 ` Vivek Goyal
2009-09-29 9:56 ` Ryo Tsuruta
2009-09-29 10:49 ` Takuya Yoshikawa
2009-09-29 14:10 ` Vivek Goyal
2009-09-29 19:53 ` Nauman Rafique
2009-09-30 8:43 ` Ryo Tsuruta
2009-09-30 11:05 ` Vivek Goyal
2009-10-01 6:41 ` Ryo Tsuruta
2009-10-01 13:31 ` Vivek Goyal
2009-10-02 2:57 ` Vivek Goyal
2009-10-02 20:27 ` Munehiro Ikeda
2009-10-05 10:38 ` Ryo Tsuruta
2009-10-05 12:31 ` Vivek Goyal
2009-10-05 14:55 ` Ryo Tsuruta
2009-10-05 17:10 ` Vivek Goyal
2009-10-05 18:11 ` Nauman Rafique
2009-10-06 7:17 ` Ryo Tsuruta
2009-10-06 11:22 ` Vivek Goyal
2009-10-07 14:38 ` Ryo Tsuruta
2009-10-07 15:09 ` Vivek Goyal
2009-10-08 2:18 ` Ryo Tsuruta
2009-10-07 16:41 ` Rik van Riel
2009-10-08 10:22 ` Ryo Tsuruta
2009-09-30 3:11 ` Vivek Goyal
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=1254551466.7716.4.camel@marge.simson.net \
--to=efault@gmx.de \
--cc=agk@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=balbir@linux.vnet.ibm.com \
--cc=containers@lists.linux-foundation.org \
--cc=dhaval@linux.vnet.ibm.com \
--cc=dm-devel@redhat.com \
--cc=dpshah@google.com \
--cc=fchecconi@gmail.com \
--cc=fernando@oss.ntt.co.jp \
--cc=jens.axboe@oracle.com \
--cc=jmarchan@redhat.com \
--cc=jmoyer@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=lizf@cn.fujitsu.com \
--cc=m-ikeda@ds.jp.nec.com \
--cc=mikew@google.com \
--cc=mingo@elte.hu \
--cc=nauman@google.com \
--cc=paolo.valente@unimore.it \
--cc=peterz@infradead.org \
--cc=riel@redhat.com \
--cc=righi.andrea@gmail.com \
--cc=ryov@valinux.co.jp \
--cc=stellplatz-nr.13a@datenparkplatz.de \
--cc=torvalds@linux-foundation.org \
--cc=vgoyal@redhat.com \
/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).