From: Christoph Lameter <clameter@sgi.com>
To: Matthew Wilcox <willy@linux.intel.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>,
Christoph Hellwig <hch@lst.de>, Mel Gorman <mel@skynet.ie>,
linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
David Chinner <dgc@sgi.com>, Jens Axboe <jens.axboe@oracle.com>,
suresh.b.siddha@intel.com
Subject: Re: SLUB performance regression vs SLAB
Date: Thu, 4 Oct 2007 12:05:35 -0700 (PDT) [thread overview]
Message-ID: <Pine.LNX.4.64.0710041156120.11927@schroedinger.engr.sgi.com> (raw)
In-Reply-To: <20071004192824.GA9852@linux.intel.com>
On Thu, 4 Oct 2007, Matthew Wilcox wrote:
> We have three runs, all with 2.6.23-rc3 plus the patches that Suresh
> applied from 20070922. The first run is with slab. The second run is
> with SLUB and the third run is SLUB plus the tuning parameters you
> recommended.
There was quite a bit of communication on tuning parameters. Guess we got
more confusion there and multiple configurations settings that I wanted to
be tested separately were merged. Setting slub_min_order to more than zero
can certainly be detrimental to performance since higher order page
allocations can cause cacheline bouncing on zone locks.
Which patches? 20070922 refers to a pull on the slab git tree on the
performance branch?
> I have a spreadsheet with Vtune data in it that was collected during
> each of these test runs, so we can see which functions are the hottest.
> I can grab that data and send it to you, if that's interesting.
Please do. Add the kernel .configs please. Is there any slab queue tuning
going on on boot with the SLAB configuration?
Include any tuning that was done to the kernel please.
> > Was the page allocator pass through patchset
> > separately applied as I requested?
>
> I don't believe so. Suresh?
If it was a git pull then the pass through was included and never taken
out.
> I think for future tests, it would be easiest if you send me a git
> reference. That way we will all know precisely what is being tested.
Sure we can do that.
> > Finally: Is there some way that I can reproduce the tests on my machines?
>
> As usual for these kinds of setups ... take a two-CPU machine, 64GB
> of memory, half a dozen fibre channel adapters, about 3000 discs,
> a commercial database, a team of experts for three months worth of
> tuning ...
>
> I don't know if anyone's tried to replicate a benchmark like this using
> Postgres. Would be nice if they have ...
Well we got our own performance test department here at SGI. If we get
them involved then we can add another 3 months until we get the test
results confirmed ;-). Seems that this is a small configuration. Why
does it take that long? And the experts knew SLAB and not SLUB right?
Lets look at all the data that you got and then see if this is enough to
figure out what is wrong.
next prev parent reply other threads:[~2007-10-04 19:05 UTC|newest]
Thread overview: 110+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-09-19 3:36 [00/17] [RFC] Virtual Compound Page Support Christoph Lameter
2007-09-19 3:36 ` [01/17] Vmalloc: Move vmalloc_to_page to mm/vmalloc Christoph Lameter
2007-09-19 3:36 ` [02/17] Vmalloc: add const Christoph Lameter
2007-09-19 3:36 ` [03/17] is_vmalloc_addr(): Check if an address is within the vmalloc boundaries Christoph Lameter
2007-09-19 6:32 ` David Rientjes
2007-09-19 7:24 ` Anton Altaparmakov
2007-09-19 8:09 ` David Rientjes
2007-09-19 8:44 ` Anton Altaparmakov
2007-09-19 9:19 ` David Rientjes
2007-09-19 13:23 ` Anton Altaparmakov
2007-09-19 17:29 ` Christoph Lameter
2007-09-19 17:52 ` Anton Altaparmakov
2007-09-19 17:29 ` Christoph Lameter
2007-09-19 17:52 ` Anton Altaparmakov
2007-09-19 3:36 ` [04/17] vmalloc: clean up page array indexing Christoph Lameter
2007-09-19 3:36 ` [05/17] vunmap: return page array Christoph Lameter
2007-09-19 8:05 ` KAMEZAWA Hiroyuki
2007-09-19 22:15 ` Christoph Lameter
2007-09-20 0:47 ` KAMEZAWA Hiroyuki
2007-09-19 3:36 ` [06/17] vmalloc_address(): Determine vmalloc address from page struct Christoph Lameter
2007-09-19 3:36 ` [07/17] GFP_VFALLBACK: Allow fallback of compound pages to virtual mappings Christoph Lameter
2007-09-19 3:36 ` [08/17] Pass vmalloc address in page->private Christoph Lameter
2007-09-19 3:36 ` [09/17] VFALLBACK: Debugging aid Christoph Lameter
2007-09-19 3:36 ` [10/17] Use GFP_VFALLBACK for sparsemem Christoph Lameter
2007-09-19 3:36 ` [11/17] GFP_VFALLBACK for zone wait table Christoph Lameter
2007-09-19 3:36 ` [12/17] Virtual Compound page allocation from interrupt context Christoph Lameter
2007-09-19 3:36 ` [13/17] Virtual compound page freeing in " Christoph Lameter
2007-09-18 20:36 ` Nick Piggin
2007-09-20 17:50 ` Christoph Lameter
2007-09-19 3:36 ` [14/17] Allow bit_waitqueue to wait on a bit in a vmalloc area Christoph Lameter
2007-09-19 4:12 ` Gabriel C
2007-09-19 17:40 ` Christoph Lameter
2007-09-19 3:36 ` [15/17] SLUB: Support virtual fallback via SLAB_VFALLBACK Christoph Lameter
2007-09-27 21:42 ` Nick Piggin
2007-09-28 17:33 ` Christoph Lameter
2007-09-28 5:14 ` Nick Piggin
2007-10-01 20:50 ` Christoph Lameter
2007-10-02 8:43 ` Nick Piggin
2007-10-04 16:16 ` SLUB performance regression vs SLAB Matthew Wilcox
2007-10-04 17:38 ` Christoph Lameter
2007-10-04 17:50 ` Arjan van de Ven
2007-10-04 17:58 ` Christoph Lameter
2007-10-04 18:26 ` Peter Zijlstra
2007-10-04 20:48 ` David Miller
2007-10-04 20:58 ` Matthew Wilcox
2007-10-04 21:05 ` David Miller
2007-10-04 21:11 ` Christoph Lameter
2007-10-04 18:32 ` Matthew Wilcox
2007-10-04 17:49 ` Christoph Lameter
2007-10-04 19:28 ` Matthew Wilcox
2007-10-04 19:05 ` Christoph Lameter [this message]
2007-10-04 19:46 ` Siddha, Suresh B
2007-10-04 20:55 ` David Miller
2007-10-04 21:02 ` Chuck Ebbert
2007-10-04 21:11 ` David Miller
2007-10-04 21:47 ` Chuck Ebbert
2007-10-04 22:07 ` David Miller
2007-10-04 22:23 ` David Chinner
2007-10-05 6:48 ` Jens Axboe
2007-10-05 9:19 ` Pekka Enberg
2007-10-05 9:28 ` Jens Axboe
2007-10-05 11:12 ` Andi Kleen
2007-10-05 12:39 ` Jens Axboe
2007-10-05 19:31 ` Christoph Lameter
2007-10-05 19:32 ` Christoph Lameter
2007-10-05 11:56 ` Matthew Wilcox
2007-10-05 12:37 ` Jens Axboe
2007-10-05 19:27 ` Christoph Lameter
2007-10-05 20:32 ` Peter Zijlstra
2007-10-05 21:31 ` David Miller
2007-10-04 21:05 ` Matthew Wilcox
2007-10-05 2:43 ` Christoph Lameter
2007-10-05 2:53 ` Arjan van de Ven
2007-09-28 17:55 ` [15/17] SLUB: Support virtual fallback via SLAB_VFALLBACK Peter Zijlstra
2007-09-28 18:20 ` Christoph Lameter
2007-09-28 18:25 ` Peter Zijlstra
2007-09-28 18:41 ` Christoph Lameter
2007-09-28 20:22 ` Nick Piggin
2007-09-28 21:14 ` Mel Gorman
2007-09-28 20:59 ` Mel Gorman
2007-09-29 8:13 ` Andrew Morton
2007-09-29 8:47 ` Peter Zijlstra
2007-09-29 8:53 ` Peter Zijlstra
2007-09-29 9:01 ` Andrew Morton
2007-09-29 9:14 ` Peter Zijlstra
2007-09-29 9:27 ` Andrew Morton
2007-09-28 20:19 ` Nick Piggin
2007-09-29 19:20 ` Andrew Morton
2007-09-29 19:09 ` Nick Piggin
2007-09-30 20:12 ` Andrew Morton
2007-09-30 4:16 ` Nick Piggin
2007-09-29 9:00 ` Andrew Morton
2007-10-01 20:55 ` Christoph Lameter
2007-10-01 21:30 ` Andrew Morton
2007-10-01 21:38 ` Christoph Lameter
2007-10-01 21:45 ` Andrew Morton
2007-10-01 21:52 ` Christoph Lameter
2007-10-02 9:19 ` Peter Zijlstra
2007-09-29 8:45 ` Peter Zijlstra
2007-10-01 21:01 ` Christoph Lameter
2007-10-02 8:37 ` Nick Piggin
2007-09-28 21:05 ` Mel Gorman
2007-10-01 21:10 ` Christoph Lameter
2007-09-19 3:36 ` [16/17] Allow virtual fallback for buffer_heads Christoph Lameter
2007-09-19 3:36 ` [17/17] Allow virtual fallback for dentries Christoph Lameter
2007-09-19 7:34 ` [00/17] [RFC] Virtual Compound Page Support Anton Altaparmakov
2007-09-19 8:34 ` Eric Dumazet
2007-09-19 17:33 ` Christoph Lameter
2007-09-19 8:24 ` Andi Kleen
2007-09-19 17:36 ` Christoph Lameter
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=Pine.LNX.4.64.0710041156120.11927@schroedinger.engr.sgi.com \
--to=clameter@sgi.com \
--cc=dgc@sgi.com \
--cc=hch@lst.de \
--cc=jens.axboe@oracle.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mel@skynet.ie \
--cc=nickpiggin@yahoo.com.au \
--cc=suresh.b.siddha@intel.com \
--cc=willy@linux.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).