From: Nishanth Aravamudan <nacc@linux.vnet.ibm.com>
To: benh@kernel.crashing.org
Cc: Fenghua Yu <fenghua.yu@intel.com>,
Tony Luck <tony.luck@intel.com>,
linux-ia64@vger.kernel.org, linux-kernel@vger.kernel.org,
linux-mm@kvack.org, David Rientjes <rientjes@google.com>,
Joonsoo Kim <iamjoonsoo.kim@lge.com>,
linuxppc-dev@lists.ozlabs.org,
Jiang Liu <jiang.liu@linux.intel.com>,
Wanpeng Li <liwanp@linux.vnet.ibm.com>
Subject: [RFC 0/2] Memoryless nodes and kworker
Date: Thu, 17 Jul 2014 16:09:23 -0700 [thread overview]
Message-ID: <20140717230923.GA32660@linux.vnet.ibm.com> (raw)
[Apologies for the large Cc list, but I believe we have the following
interested parties:
x86 (recently posted memoryless node support)
ia64 (existing memoryless node support)
ppc (existing memoryless node support)
previous discussion of how to solve Anton's issue with slab usage
workqueue contributors/maintainers]
There is an issue currently where NUMA information is used on powerpc
(and possibly ia64) before it has been read from the device-tree, which
leads to large slab consumption with CONFIG_SLUB and memoryless nodes.
While testing memoryless nodes on PowerKVM guests with the patches in
this series, with a guest topology of
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
node 0 size: 0 MB
node 0 free: 0 MB
node 1 cpus: 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
node 1 size: 16336 MB
node 1 free: 15329 MB
node distances:
node 0 1
0: 10 40
1: 40 10
the slab consumption decreases from
Slab: 932416 kB
SUnreclaim: 902336 kB
to
Slab: 395264 kB
SUnreclaim: 359424 kB
And we see a corresponding increase in the slab efficiency from
slab mem objs slabs
used active active
------------------------------------------------------------
kmalloc-16384 337 MB 11.28% 100.00%
task_struct 288 MB 9.93% 100.00%
to
slab mem objs slabs
used active active
------------------------------------------------------------
kmalloc-16384 37 MB 100.00% 100.00%
task_struct 31 MB 100.00% 100.00%
It turns out we see this large slab usage due to using the wrong NUMA
information when creating kthreads.
Two changes are required, one of which is in the workqueue code and one
of which is in the powerpc initialization. Note that ia64 may want to
consider something similar.
next reply other threads:[~2014-07-17 23:09 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-07-17 23:09 Nishanth Aravamudan [this message]
2014-07-17 23:09 ` [RFC 1/2] workqueue: use the nearest NUMA node, not the local one Nishanth Aravamudan
2014-07-17 23:15 ` [RFC 2/2] powerpc: reorder per-cpu NUMA information's initialization Nishanth Aravamudan
2014-07-18 8:11 ` [RFC 1/2] workqueue: use the nearest NUMA node, not the local one Lai Jiangshan
2014-07-18 17:33 ` Nish Aravamudan
2014-07-18 11:20 ` [RFC 0/2] Memoryless nodes and kworker Tejun Heo
2014-07-18 17:42 ` Nish Aravamudan
2014-07-18 18:00 ` Tejun Heo
2014-07-18 18:01 ` Tejun Heo
2014-07-18 18:12 ` Nish Aravamudan
2014-07-18 18:19 ` Tejun Heo
2014-07-18 18:47 ` Nish Aravamudan
2014-07-18 18:58 ` Tejun Heo
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=20140717230923.GA32660@linux.vnet.ibm.com \
--to=nacc@linux.vnet.ibm.com \
--cc=benh@kernel.crashing.org \
--cc=fenghua.yu@intel.com \
--cc=iamjoonsoo.kim@lge.com \
--cc=jiang.liu@linux.intel.com \
--cc=linux-ia64@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=liwanp@linux.vnet.ibm.com \
--cc=rientjes@google.com \
--cc=tony.luck@intel.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).