All of lore.kernel.org
 help / color / mirror / Atom feed
From: Marcelo Tosatti <mtosatti-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
To: "Leonardo Brás" <leobras-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
Cc: Michal Hocko <mhocko-IBi9RG/b67k@public.gmane.org>,
	Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>,
	Roman Gushchin
	<roman.gushchin-fxUVXftIFDnyG1zEObXtfA@public.gmane.org>,
	Shakeel Butt <shakeelb-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>,
	Muchun Song <muchun.song-fxUVXftIFDnyG1zEObXtfA@public.gmane.org>,
	Andrew Morton
	<akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>,
	cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
	linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: Re: [PATCH v2 0/5] Introduce memcg_stock_pcp remote draining
Date: Wed, 25 Jan 2023 15:22:00 -0300	[thread overview]
Message-ID: <Y9FzSBw10MGXm2TK@tpad> (raw)
In-Reply-To: <9e61ab53e1419a144f774b95230b789244895424.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>

On Wed, Jan 25, 2023 at 08:06:46AM -0300, Leonardo Brás wrote:
> On Wed, 2023-01-25 at 09:33 +0100, Michal Hocko wrote:
> > On Wed 25-01-23 04:34:57, Leonardo Bras wrote:
> > > Disclaimer:
> > > a - The cover letter got bigger than expected, so I had to split it in
> > >     sections to better organize myself. I am not very confortable with it.
> > > b - Performance numbers below did not include patch 5/5 (Remove flags
> > >     from memcg_stock_pcp), which could further improve performance for
> > >     drain_all_stock(), but I could only notice the optimization at the
> > >     last minute.
> > > 
> > > 
> > > 0 - Motivation:
> > > On current codebase, when drain_all_stock() is ran, it will schedule a
> > > drain_local_stock() for each cpu that has a percpu stock associated with a
> > > descendant of a given root_memcg.
> > > 
> > > This happens even on 'isolated cpus', a feature commonly used on workloads that
> > > are sensitive to interruption and context switching such as vRAN and Industrial
> > > Control Systems.
> > > 
> > > Since this scheduling behavior is a problem to those workloads, the proposal is
> > > to replace the current local_lock + schedule_work_on() solution with a per-cpu
> > > spinlock.
> > 
> > If IIRC we have also discussed that isolated CPUs can simply opt out
> > from the pcp caching and therefore the problem would be avoided
> > altogether without changes to the locking scheme. I do not see anything
> > regarding that in this submission. Could you elaborate why you have
> > abandoned this option?
> 
> Hello Michal,
> 
> I understand pcp caching is a nice to have.
> So while I kept the idea of disabling pcp caching in mind as an option, I first
> tried to understand what kind of impacts we would be seeing when trying to
> change the locking scheme.

Remote draining reduces interruptions whether CPU 
is marked as isolated or not:

- Allows isolated CPUs from benefiting of pcp caching.
- Removes the interruption to non isolated CPUs. See for example 

https://lkml.org/lkml/2022/6/13/2769

"Minchan Kim tested this independently and reported;

       My workload is not NOHZ CPUs but run apps under heavy memory
       pressure so they goes to direct reclaim and be stuck on
       drain_all_pages until work on workqueue run.

       unit: nanosecond
       max(dur)        avg(dur)                count(dur)
       166713013       487511.77786438033      1283

       From traces, system encountered the drain_all_pages 1283 times and
       worst case was 166ms and avg was 487us.

       The other problem was alloc_contig_range in CMA. The PCP draining
       takes several hundred millisecond sometimes though there is no
       memory pressure or a few of pages to be migrated out but CPU were
       fully booked.

       Your patch perfectly removed those wasted time."


> After I raised the data in the cover letter, I found that the performance impact
> appears not be that big. So in order to try keeping the pcp cache on isolated
> cpus active, I decided to focus effort on the locking scheme change.
> 
> I mean, my rationale is: if is there a non-expensive way of keeping the feature,
> why should we abandon it?
> 
> Best regards,
> Leo
> 
> 
> 
> 
> 
> 
> 
> 


WARNING: multiple messages have this Message-ID (diff)
From: Marcelo Tosatti <mtosatti@redhat.com>
To: "Leonardo Brás" <leobras@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Roman Gushchin <roman.gushchin@linux.dev>,
	Shakeel Butt <shakeelb@google.com>,
	Muchun Song <muchun.song@linux.dev>,
	Andrew Morton <akpm@linux-foundation.org>,
	cgroups@vger.kernel.org, linux-mm@kvack.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 0/5] Introduce memcg_stock_pcp remote draining
Date: Wed, 25 Jan 2023 15:22:00 -0300	[thread overview]
Message-ID: <Y9FzSBw10MGXm2TK@tpad> (raw)
In-Reply-To: <9e61ab53e1419a144f774b95230b789244895424.camel@redhat.com>

On Wed, Jan 25, 2023 at 08:06:46AM -0300, Leonardo Brás wrote:
> On Wed, 2023-01-25 at 09:33 +0100, Michal Hocko wrote:
> > On Wed 25-01-23 04:34:57, Leonardo Bras wrote:
> > > Disclaimer:
> > > a - The cover letter got bigger than expected, so I had to split it in
> > >     sections to better organize myself. I am not very confortable with it.
> > > b - Performance numbers below did not include patch 5/5 (Remove flags
> > >     from memcg_stock_pcp), which could further improve performance for
> > >     drain_all_stock(), but I could only notice the optimization at the
> > >     last minute.
> > > 
> > > 
> > > 0 - Motivation:
> > > On current codebase, when drain_all_stock() is ran, it will schedule a
> > > drain_local_stock() for each cpu that has a percpu stock associated with a
> > > descendant of a given root_memcg.
> > > 
> > > This happens even on 'isolated cpus', a feature commonly used on workloads that
> > > are sensitive to interruption and context switching such as vRAN and Industrial
> > > Control Systems.
> > > 
> > > Since this scheduling behavior is a problem to those workloads, the proposal is
> > > to replace the current local_lock + schedule_work_on() solution with a per-cpu
> > > spinlock.
> > 
> > If IIRC we have also discussed that isolated CPUs can simply opt out
> > from the pcp caching and therefore the problem would be avoided
> > altogether without changes to the locking scheme. I do not see anything
> > regarding that in this submission. Could you elaborate why you have
> > abandoned this option?
> 
> Hello Michal,
> 
> I understand pcp caching is a nice to have.
> So while I kept the idea of disabling pcp caching in mind as an option, I first
> tried to understand what kind of impacts we would be seeing when trying to
> change the locking scheme.

Remote draining reduces interruptions whether CPU 
is marked as isolated or not:

- Allows isolated CPUs from benefiting of pcp caching.
- Removes the interruption to non isolated CPUs. See for example 

https://lkml.org/lkml/2022/6/13/2769

"Minchan Kim tested this independently and reported;

       My workload is not NOHZ CPUs but run apps under heavy memory
       pressure so they goes to direct reclaim and be stuck on
       drain_all_pages until work on workqueue run.

       unit: nanosecond
       max(dur)        avg(dur)                count(dur)
       166713013       487511.77786438033      1283

       From traces, system encountered the drain_all_pages 1283 times and
       worst case was 166ms and avg was 487us.

       The other problem was alloc_contig_range in CMA. The PCP draining
       takes several hundred millisecond sometimes though there is no
       memory pressure or a few of pages to be migrated out but CPU were
       fully booked.

       Your patch perfectly removed those wasted time."


> After I raised the data in the cover letter, I found that the performance impact
> appears not be that big. So in order to try keeping the pcp cache on isolated
> cpus active, I decided to focus effort on the locking scheme change.
> 
> I mean, my rationale is: if is there a non-expensive way of keeping the feature,
> why should we abandon it?
> 
> Best regards,
> Leo
> 
> 
> 
> 
> 
> 
> 
> 



  parent reply	other threads:[~2023-01-25 18:22 UTC|newest]

Thread overview: 85+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-01-25  7:34 [PATCH v2 0/5] Introduce memcg_stock_pcp remote draining Leonardo Bras
2023-01-25  7:34 ` Leonardo Bras
     [not found] ` <20230125073502.743446-1-leobras-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-01-25  7:34   ` [PATCH v2 1/5] mm/memcontrol: Align percpu memcg_stock to cache Leonardo Bras
2023-01-25  7:34     ` Leonardo Bras
2023-01-25  7:34   ` [PATCH v2 2/5] mm/memcontrol: Change stock_lock type from local_lock_t to spinlock_t Leonardo Bras
2023-01-25  7:34     ` Leonardo Bras
2023-01-25  7:35   ` [PATCH v2 3/5] mm/memcontrol: Reorder memcg_stock_pcp members to avoid holes Leonardo Bras
2023-01-25  7:35     ` Leonardo Bras
2023-01-25  7:35   ` [PATCH v2 5/5] mm/memcontrol: Remove flags from memcg_stock_pcp Leonardo Bras
2023-01-25  7:35     ` Leonardo Bras
2023-01-25  8:33   ` [PATCH v2 0/5] Introduce memcg_stock_pcp remote draining Michal Hocko
2023-01-25  8:33     ` Michal Hocko
     [not found]     ` <Y9DpbVF+JR/G+5Or-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-25 11:06       ` Leonardo Brás
2023-01-25 11:06         ` Leonardo Brás
     [not found]         ` <9e61ab53e1419a144f774b95230b789244895424.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-01-25 11:39           ` Michal Hocko
2023-01-25 11:39             ` Michal Hocko
2023-01-25 18:22           ` Marcelo Tosatti [this message]
2023-01-25 18:22             ` Marcelo Tosatti
2023-01-25 23:14             ` Roman Gushchin
2023-01-25 23:14               ` Roman Gushchin
     [not found]               ` <Y9G36AiqPPFDlax3-+xijCwNIfdoLQcUKs7qKB+WAnPUfkyWGUBSOeVevoDU@public.gmane.org>
2023-01-26  7:41                 ` Michal Hocko
2023-01-26  7:41                   ` Michal Hocko
2023-01-26 18:03                   ` Marcelo Tosatti
2023-01-26 19:20                     ` Michal Hocko
     [not found]                       ` <Y9LSjnNEEUiF/70R-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-27  0:32                         ` Marcelo Tosatti
2023-01-27  0:32                           ` Marcelo Tosatti
2023-01-27  6:58                           ` Michal Hocko
2023-01-27  6:58                             ` Michal Hocko
2023-02-01 18:31                       ` Roman Gushchin
2023-01-26 23:12                   ` Roman Gushchin
     [not found]                     ` <Y9MI42NSLooyVZNu-+xijCwNIfdoLQcUKs7qKB+WAnPUfkyWGUBSOeVevoDU@public.gmane.org>
2023-01-27  7:11                       ` Michal Hocko
2023-01-27  7:11                         ` Michal Hocko
2023-01-27  7:22                         ` Leonardo Brás
     [not found]                           ` <52a0f1e593b1ec0ca7e417ba37680d65df22de82.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-01-27  8:12                             ` Leonardo Brás
2023-01-27  8:12                               ` Leonardo Brás
     [not found]                               ` <601fc35a8cc2167e53e45c636fccb2d899fd7c50.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-01-27  9:23                                 ` Michal Hocko
2023-01-27  9:23                                   ` Michal Hocko
2023-01-27 13:03                                 ` Frederic Weisbecker
2023-01-27 13:03                                   ` Frederic Weisbecker
     [not found]                         ` <Y9N5CI8PpsfiaY9c-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-27 13:58                           ` Michal Hocko
2023-01-27 13:58                             ` Michal Hocko
     [not found]                             ` <Y9PYe1X7dRQOcahg-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-27 18:18                               ` Roman Gushchin
2023-01-27 18:18                                 ` Roman Gushchin
     [not found]                                 ` <Y9QVWwAreTlDVdZ0-+xijCwNIfdoLQcUKs7qKB+WAnPUfkyWGUBSOeVevoDU@public.gmane.org>
2023-02-03 15:21                                   ` Michal Hocko
2023-02-03 15:21                                     ` Michal Hocko
     [not found]                                     ` <Y90mZQhW89HtYfT9-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-02-03 19:25                                       ` Roman Gushchin
2023-02-03 19:25                                         ` Roman Gushchin
     [not found]                                         ` <Y91fnF5uEcSA0/99-+xijCwNIfdoLQcUKs7qKB+WAnPUfkyWGUBSOeVevoDU@public.gmane.org>
2023-02-13 13:36                                           ` Michal Hocko
2023-02-13 13:36                                             ` Michal Hocko
2023-01-27  7:14                       ` Leonardo Brás
2023-01-27  7:14                         ` Leonardo Brás
     [not found]                         ` <55ac6e3cbb97c7d13c49c3125c1455d8a2c785c3.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-01-27  7:20                           ` Michal Hocko
2023-01-27  7:20                             ` Michal Hocko
2023-01-27  7:35                             ` Leonardo Brás
2023-01-27  9:29                               ` Michal Hocko
     [not found]                                 ` <Y9OZezjUPITtEvTx-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-27 19:29                                   ` Leonardo Brás
2023-01-27 19:29                                     ` Leonardo Brás
2023-01-27 23:50                                     ` Roman Gushchin
2023-01-26 18:19                 ` Marcelo Tosatti
2023-01-26 18:19                   ` Marcelo Tosatti
2023-01-27  5:40                   ` Leonardo Brás
2023-01-27  5:40                     ` Leonardo Brás
2023-01-26  2:01             ` Hillf Danton
2023-01-26  7:45             ` Michal Hocko
2023-01-26  7:45               ` Michal Hocko
     [not found]               ` <Y9IvoDJbLbFcitTc-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-26 18:14                 ` Marcelo Tosatti
2023-01-26 18:14                   ` Marcelo Tosatti
2023-01-26 19:13                   ` Michal Hocko
2023-01-26 19:13                     ` Michal Hocko
     [not found]                     ` <Y9LQ615H13RmG7wL-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-01-27  6:55                       ` Leonardo Brás
2023-01-27  6:55                         ` Leonardo Brás
     [not found]                         ` <0122005439ffb7895efda7a1a67992cbe41392fe.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-01-31 11:35                           ` Marcelo Tosatti
2023-01-31 11:35                             ` Marcelo Tosatti
2023-02-01  4:36                             ` Leonardo Brás
     [not found]                               ` <5ba79c4feb829ed75cfd98cf5c8042dcb2ddea91.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-02-01 12:52                                 ` Michal Hocko
2023-02-01 12:52                                   ` Michal Hocko
2023-02-01 12:41                             ` Michal Hocko
2023-02-01 12:41                               ` Michal Hocko
     [not found]                               ` <Y9pd7AxAILUSHrpe-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2023-02-04  4:55                                 ` Leonardo Brás
2023-02-04  4:55                                   ` Leonardo Brás
2023-02-05 19:49                                   ` Roman Gushchin
2023-02-07  3:18                                     ` Leonardo Brás
     [not found]                                       ` <4b232f47e038ab6fcaa0114f73c28d4bf8799f84.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2023-02-08 19:23                                         ` Roman Gushchin
2023-02-08 19:23                                           ` Roman Gushchin
2023-01-25  7:35 ` [PATCH v2 4/5] mm/memcontrol: Perform all stock drain in current CPU Leonardo Bras

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=Y9FzSBw10MGXm2TK@tpad \
    --to=mtosatti-h+wxahxf7alqt0dzr+alfa@public.gmane.org \
    --cc=akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org \
    --cc=cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org \
    --cc=leobras-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
    --cc=linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org \
    --cc=mhocko-IBi9RG/b67k@public.gmane.org \
    --cc=muchun.song-fxUVXftIFDnyG1zEObXtfA@public.gmane.org \
    --cc=roman.gushchin-fxUVXftIFDnyG1zEObXtfA@public.gmane.org \
    --cc=shakeelb-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.