From: Ingo Molnar <mingo@kernel.org>
To: Mel Gorman <mgorman@suse.de>
Cc: Dave Chinner <david@fromorbit.com>,
Andrew Morton <akpm@linux-foundation.org>,
Linus Torvalds <torvalds@linux-foundation.org>,
Aneesh Kumar <aneesh.kumar@linux.vnet.ibm.com>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Linux-MM <linux-mm@kvack.org>,
xfs@oss.sgi.com, linuxppc-dev@lists.ozlabs.org
Subject: Re: [PATCH 4/4] mm: numa: Slow PTE scan rate if migration failures occur
Date: Sat, 7 Mar 2015 17:36:58 +0100 [thread overview]
Message-ID: <20150307163657.GA9702@gmail.com> (raw)
In-Reply-To: <1425741651-29152-5-git-send-email-mgorman@suse.de>
* Mel Gorman <mgorman@suse.de> wrote:
> Dave Chinner reported the following on https://lkml.org/lkml/2015/3/1/226
>
> Across the board the 4.0-rc1 numbers are much slower, and the
> degradation is far worse when using the large memory footprint
> configs. Perf points straight at the cause - this is from 4.0-rc1 on
> the "-o bhash=101073" config:
>
> [...]
> 4.0.0-rc1 4.0.0-rc1 3.19.0
> vanilla slowscan-v2 vanilla
> User 53384.29 56093.11 46119.12
> System 692.14 311.64 306.41
> Elapsed 1236.87 1328.61 1039.88
>
> Note that the system CPU usage is now similar to 3.19-vanilla.
Similar, but still worse, and also the elapsed time is still much
worse. User time is much higher, although it's the same amount of work
done on every kernel, right?
> I also tested with a workload very similar to Dave's. The machine
> configuration and storage is completely different so it's not an
> equivalent test unfortunately. It's reporting the elapsed time and
> CPU time while fsmark is running to create the inodes and when
> runnig xfsrepair afterwards
>
> xfsrepair
> 4.0.0-rc1 4.0.0-rc1 3.19.0
> vanilla slowscan-v2 vanilla
> Min real-fsmark 1157.41 ( 0.00%) 1150.38 ( 0.61%) 1164.44 ( -0.61%)
> Min syst-fsmark 3998.06 ( 0.00%) 3988.42 ( 0.24%) 4016.12 ( -0.45%)
> Min real-xfsrepair 497.64 ( 0.00%) 456.87 ( 8.19%) 442.64 ( 11.05%)
> Min syst-xfsrepair 500.61 ( 0.00%) 263.41 ( 47.38%) 194.97 ( 61.05%)
> Amean real-fsmark 1166.63 ( 0.00%) 1155.97 ( 0.91%) 1166.28 ( 0.03%)
> Amean syst-fsmark 4020.94 ( 0.00%) 4004.19 ( 0.42%) 4025.87 ( -0.12%)
> Amean real-xfsrepair 507.85 ( 0.00%) 459.58 ( 9.50%) 447.66 ( 11.85%)
> Amean syst-xfsrepair 519.88 ( 0.00%) 281.63 ( 45.83%) 202.93 ( 60.97%)
> Stddev real-fsmark 6.55 ( 0.00%) 3.97 ( 39.30%) 1.44 ( 77.98%)
> Stddev syst-fsmark 16.22 ( 0.00%) 15.09 ( 6.96%) 9.76 ( 39.86%)
> Stddev real-xfsrepair 11.17 ( 0.00%) 3.41 ( 69.43%) 5.57 ( 50.17%)
> Stddev syst-xfsrepair 13.98 ( 0.00%) 19.94 (-42.60%) 5.69 ( 59.31%)
> CoeffVar real-fsmark 0.56 ( 0.00%) 0.34 ( 38.74%) 0.12 ( 77.97%)
> CoeffVar syst-fsmark 0.40 ( 0.00%) 0.38 ( 6.57%) 0.24 ( 39.93%)
> CoeffVar real-xfsrepair 2.20 ( 0.00%) 0.74 ( 66.22%) 1.24 ( 43.47%)
> CoeffVar syst-xfsrepair 2.69 ( 0.00%) 7.08 (-163.23%) 2.80 ( -4.23%)
> Max real-fsmark 1171.98 ( 0.00%) 1159.25 ( 1.09%) 1167.96 ( 0.34%)
> Max syst-fsmark 4033.84 ( 0.00%) 4024.53 ( 0.23%) 4039.20 ( -0.13%)
> Max real-xfsrepair 523.40 ( 0.00%) 464.40 ( 11.27%) 455.42 ( 12.99%)
> Max syst-xfsrepair 533.37 ( 0.00%) 309.38 ( 42.00%) 207.94 ( 61.01%)
>
> The key point is that system CPU usage for xfsrepair (syst-xfsrepair)
> is almost cut in half. It's still not as low as 3.19-vanilla but it's
> much closer
>
> 4.0.0-rc1 4.0.0-rc1 3.19.0
> vanilla slowscan-v2 vanilla
> NUMA alloc hit 146138883 121929782 104019526
> NUMA alloc miss 13146328 11456356 7806370
> NUMA interleave hit 0 0 0
> NUMA alloc local 146060848 121865921 103953085
> NUMA base PTE updates 242201535 117237258 216624143
> NUMA huge PMD updates 113270 52121 127782
> NUMA page range updates 300195775 143923210 282048527
> NUMA hint faults 180388025 87299060 147235021
> NUMA hint local faults 72784532 32939258 61866265
> NUMA hint local percent 40 37 42
> NUMA pages migrated 71175262 41395302 23237799
>
> Note the big differences in faults trapped and pages migrated.
> 3.19-vanilla still migrated fewer pages but if necessary the
> threshold at which we start throttling migrations can be lowered.
This too is still worse than what v3.19 had.
So what worries me is that Dave bisected the regression to:
4d9424669946 ("mm: convert p[te|md]_mknonnuma and remaining page table manipulations")
And clearly your patch #4 just tunes balancing/migration intensity -
is that a workaround for the real problem/bug?
And the patch Dave bisected to is a relatively simple patch.
Why not simply revert it to see whether that cures much of the
problem?
Am I missing something fundamental?
Thanks,
Ingo
next prev parent reply other threads:[~2015-03-07 16:37 UTC|newest]
Thread overview: 49+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-03-07 15:20 [RFC PATCH 0/4] Automatic NUMA balancing and PROT_NONE handling followup v2r8 Mel Gorman
2015-03-07 15:20 ` [PATCH 1/4] mm: thp: Return the correct value for change_huge_pmd Mel Gorman
2015-03-07 20:13 ` Linus Torvalds
2015-03-07 20:31 ` Linus Torvalds
2015-03-07 20:56 ` Mel Gorman
2015-03-07 15:20 ` [PATCH 2/4] mm: numa: Remove migrate_ratelimited Mel Gorman
2015-03-07 15:20 ` [PATCH 3/4] mm: numa: Mark huge PTEs young when clearing NUMA hinting faults Mel Gorman
2015-03-07 18:33 ` Linus Torvalds
2015-03-07 18:42 ` Linus Torvalds
2015-03-07 15:20 ` [PATCH 4/4] mm: numa: Slow PTE scan rate if migration failures occur Mel Gorman
2015-03-07 16:36 ` Ingo Molnar [this message]
2015-03-07 17:37 ` Mel Gorman
2015-03-08 9:54 ` Ingo Molnar
2015-03-07 19:12 ` Linus Torvalds
2015-03-08 10:02 ` Ingo Molnar
2015-03-08 18:35 ` Linus Torvalds
2015-03-08 18:46 ` Linus Torvalds
2015-03-09 11:29 ` Dave Chinner
2015-03-09 16:52 ` Linus Torvalds
2015-03-09 19:19 ` Dave Chinner
2015-03-10 23:55 ` Linus Torvalds
2015-03-12 13:10 ` Mel Gorman
2015-03-12 16:20 ` Linus Torvalds
2015-03-12 18:49 ` Mel Gorman
2015-03-17 7:06 ` Dave Chinner
2015-03-17 16:53 ` Linus Torvalds
2015-03-17 20:51 ` Dave Chinner
2015-03-17 21:30 ` Linus Torvalds
2015-03-17 22:08 ` Dave Chinner
2015-03-18 16:08 ` Linus Torvalds
2015-03-18 17:31 ` Linus Torvalds
2015-03-18 22:23 ` Dave Chinner
2015-03-19 14:10 ` Mel Gorman
2015-03-19 18:09 ` Linus Torvalds
2015-03-19 21:41 ` Linus Torvalds
2015-03-19 22:41 ` Dave Chinner
2015-03-19 23:05 ` Linus Torvalds
2015-03-19 23:23 ` Dave Chinner
2015-03-20 0:23 ` Dave Chinner
2015-03-20 1:29 ` Linus Torvalds
2015-03-20 4:13 ` Dave Chinner
2015-03-20 17:02 ` Linus Torvalds
2015-03-23 12:01 ` Mel Gorman
2015-03-20 10:12 ` Mel Gorman
2015-03-20 9:56 ` Mel Gorman
2015-03-08 20:40 ` Mel Gorman
2015-03-09 21:02 ` Mel Gorman
2015-03-10 13:08 ` Mel Gorman
2015-03-08 9:41 ` 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=20150307163657.GA9702@gmail.com \
--to=mingo@kernel.org \
--cc=akpm@linux-foundation.org \
--cc=aneesh.kumar@linux.vnet.ibm.com \
--cc=david@fromorbit.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=mgorman@suse.de \
--cc=torvalds@linux-foundation.org \
--cc=xfs@oss.sgi.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).