From: Wu Fengguang <wfg@linux.intel.com>
To: Vivek Goyal <vgoyal@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
Shaohua Li <shaohua.li@intel.com>,
lkml <linux-kernel@vger.kernel.org>,
linux-mm <linux-mm@kvack.org>, Jens Axboe <axboe@kernel.dk>,
Herbert Poetzl <herbert@13thfloor.at>,
Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: [PATCH] fix readahead pipeline break caused by block plug
Date: Wed, 1 Feb 2012 15:10:00 +0800 [thread overview]
Message-ID: <20120201071000.GB29083@localhost> (raw)
In-Reply-To: <20120201033653.GA12092@redhat.com>
On Tue, Jan 31, 2012 at 10:36:53PM -0500, Vivek Goyal wrote:
> On Tue, Jan 31, 2012 at 05:22:17PM -0500, Vivek Goyal wrote:
> [..]
>
> > >
> > > We've never really bothered making the /dev/sda[X] I/O very efficient
> > > for large I/O's under the (probably wrong) assumption that it isn't a
> > > very interesting case. Regular files will (or should) use the mpage
> > > functions, via address_space_operations.readpages(). fs/blockdev.c
> > > doesn't even implement it.
> > >
> > > > and by the time all the pages
> > > > are submitted and one big merged request is formed it wates lot of time.
> > >
> > > But that was the case in eariler kernels too. Why did it change?
> >
> > Actually, I assumed that the case of reading /dev/sda[X] worked well in
> > earlier kernels. Sorry about that. Will build a 2.6.38 kernel tonight
> > and run the test case again to make sure we had same overhead and
> > relatively poor performance while reading /dev/sda[X].
>
> Ok, I tried it with 2.6.38 kernel and results look more or less same.
> Throughput varied between 105MB to 145MB. Many a times it was close to
> 110MB and other times it was 145MB. Don't know what causes that spike
> sometimes.
The block device really has some aged performance bug. Which
interestingly only show up in some test environments...
> I still see that IO is being submitted one page at a time. The only
> real difference seems to be that queue unplug happening at random times
> and many a times we are submitting much smaller requests (40 sectors, 48
> sectors etc).
Would you share the blktrace data?
Thanks,
Fengguang
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
WARNING: multiple messages have this Message-ID (diff)
From: Wu Fengguang <wfg@linux.intel.com>
To: Vivek Goyal <vgoyal@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
Shaohua Li <shaohua.li@intel.com>,
lkml <linux-kernel@vger.kernel.org>,
linux-mm <linux-mm@kvack.org>, Jens Axboe <axboe@kernel.dk>,
Herbert Poetzl <herbert@13thfloor.at>,
Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: [PATCH] fix readahead pipeline break caused by block plug
Date: Wed, 1 Feb 2012 15:10:00 +0800 [thread overview]
Message-ID: <20120201071000.GB29083@localhost> (raw)
In-Reply-To: <20120201033653.GA12092@redhat.com>
On Tue, Jan 31, 2012 at 10:36:53PM -0500, Vivek Goyal wrote:
> On Tue, Jan 31, 2012 at 05:22:17PM -0500, Vivek Goyal wrote:
> [..]
>
> > >
> > > We've never really bothered making the /dev/sda[X] I/O very efficient
> > > for large I/O's under the (probably wrong) assumption that it isn't a
> > > very interesting case. Regular files will (or should) use the mpage
> > > functions, via address_space_operations.readpages(). fs/blockdev.c
> > > doesn't even implement it.
> > >
> > > > and by the time all the pages
> > > > are submitted and one big merged request is formed it wates lot of time.
> > >
> > > But that was the case in eariler kernels too. Why did it change?
> >
> > Actually, I assumed that the case of reading /dev/sda[X] worked well in
> > earlier kernels. Sorry about that. Will build a 2.6.38 kernel tonight
> > and run the test case again to make sure we had same overhead and
> > relatively poor performance while reading /dev/sda[X].
>
> Ok, I tried it with 2.6.38 kernel and results look more or less same.
> Throughput varied between 105MB to 145MB. Many a times it was close to
> 110MB and other times it was 145MB. Don't know what causes that spike
> sometimes.
The block device really has some aged performance bug. Which
interestingly only show up in some test environments...
> I still see that IO is being submitted one page at a time. The only
> real difference seems to be that queue unplug happening at random times
> and many a times we are submitting much smaller requests (40 sectors, 48
> sectors etc).
Would you share the blktrace data?
Thanks,
Fengguang
next prev parent reply other threads:[~2012-02-01 7:20 UTC|newest]
Thread overview: 56+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-01-31 7:59 [PATCH] fix readahead pipeline break caused by block plug Shaohua Li
2012-01-31 7:59 ` Shaohua Li
2012-01-31 8:36 ` Christoph Hellwig
2012-01-31 8:36 ` Christoph Hellwig
2012-01-31 8:48 ` Eric Dumazet
2012-01-31 8:48 ` Eric Dumazet
2012-01-31 8:50 ` Herbert Poetzl
2012-01-31 8:50 ` Herbert Poetzl
2012-01-31 8:53 ` Shaohua Li
2012-01-31 8:53 ` Shaohua Li
2012-01-31 9:17 ` Eric Dumazet
2012-01-31 9:17 ` Eric Dumazet
2012-01-31 10:20 ` Wu Fengguang
2012-01-31 10:20 ` Wu Fengguang
2012-01-31 10:34 ` Wu Fengguang
2012-01-31 10:34 ` Wu Fengguang
2012-01-31 10:46 ` Christoph Hellwig
2012-01-31 10:46 ` Christoph Hellwig
2012-01-31 10:57 ` Wu Fengguang
2012-01-31 10:57 ` Wu Fengguang
2012-01-31 11:34 ` Christoph Hellwig
2012-01-31 11:34 ` Christoph Hellwig
2012-01-31 11:42 ` Wu Fengguang
2012-01-31 11:42 ` Wu Fengguang
2012-01-31 11:57 ` Christoph Hellwig
2012-01-31 11:57 ` Christoph Hellwig
2012-01-31 12:20 ` Wu Fengguang
2012-01-31 12:20 ` Wu Fengguang
2012-02-01 2:25 ` Shaohua Li
2012-02-01 2:25 ` Shaohua Li
2012-01-31 14:47 ` Vivek Goyal
2012-01-31 14:47 ` Vivek Goyal
2012-01-31 20:23 ` Vivek Goyal
2012-01-31 20:23 ` Vivek Goyal
2012-01-31 22:03 ` Vivek Goyal
2012-01-31 22:03 ` Vivek Goyal
2012-01-31 22:13 ` Andrew Morton
2012-01-31 22:13 ` Andrew Morton
2012-01-31 22:22 ` Vivek Goyal
2012-01-31 22:22 ` Vivek Goyal
2012-02-01 3:36 ` Vivek Goyal
2012-02-01 3:36 ` Vivek Goyal
2012-02-01 7:10 ` Wu Fengguang [this message]
2012-02-01 7:10 ` Wu Fengguang
2012-02-01 16:01 ` Vivek Goyal
2012-02-01 16:01 ` Vivek Goyal
2012-02-01 9:18 ` Christoph Hellwig
2012-02-01 9:18 ` Christoph Hellwig
2012-02-01 20:10 ` Vivek Goyal
2012-02-01 20:10 ` Vivek Goyal
2012-02-01 20:13 ` Jeff Moyer
2012-02-01 20:13 ` Jeff Moyer
2012-02-01 20:22 ` Andrew Morton
2012-02-01 20:22 ` Andrew Morton
2012-02-01 7:02 ` Wu Fengguang
2012-02-01 7:02 ` Wu Fengguang
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=20120201071000.GB29083@localhost \
--to=wfg@linux.intel.com \
--cc=akpm@linux-foundation.org \
--cc=axboe@kernel.dk \
--cc=eric.dumazet@gmail.com \
--cc=herbert@13thfloor.at \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=shaohua.li@intel.com \
--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 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.