All of lore.kernel.org
 help / color / mirror / Atom feed
From: Mark Nelson <mnelson@redhat.com>
To: Sage Weil <sage@newdream.net>
Cc: Somnath Roy <Somnath.Roy@sandisk.com>,
	"Chen, Xiaoxi" <xiaoxi.chen@intel.com>,
	Haomai Wang <haomaiwang@gmail.com>,
	ceph-devel <ceph-devel@vger.kernel.org>
Subject: Re: Regarding newstore performance
Date: Fri, 17 Apr 2015 22:34:09 -0500	[thread overview]
Message-ID: <5531D0B1.1070406@redhat.com> (raw)
In-Reply-To: <alpine.DEB.2.00.1504170844161.18547@cobra.newdream.net>

[-- Attachment #1: Type: text/plain, Size: 945 bytes --]

On 04/17/2015 10:46 AM, Sage Weil wrote:
> On Fri, 17 Apr 2015, Mark Nelson wrote:
>> On 04/16/2015 07:38 PM, Sage Weil wrote:
>> Absolutely!  I'll get some tests running now.  Looks like everyone is jumping
>> on the libaio bandwagon which naively seems like the right way to me too.  Can
>> you talk a little bit more about how you'd see fdatasync work in this case
>> though vs the threaded implementation?
>
> That I'm not certain about, not sure if I need O_DSYNC or if the libaio
> fsync hook actually works; the docs are ambiguous.

Ran tests with 8 WAL threads and compared to filestore numbers.  This is 
using the Xiaoxi tunables.  Things are definitely looking better on 
spinning disks.  Random writes even without the SSD WAL are far more 
comparable to filestore than previously.  Sequential writes are better 
but still sometimes quite a bit worse.

PCIe SSD numbers have improved marginally, but still far below filestore.

Mark

[-- Attachment #2: newstore_wal_threads.pdf --]
[-- Type: application/pdf, Size: 72069 bytes --]

  reply	other threads:[~2015-04-18  3:34 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-04-15  6:01 Regarding newstore performance Somnath Roy
2015-04-15 12:23 ` Haomai Wang
2015-04-15 16:07   ` Somnath Roy
2015-04-16  1:47     ` Chen, Xiaoxi
2015-04-16  4:22       ` Somnath Roy
2015-04-16  6:17         ` Somnath Roy
2015-04-16 18:17           ` Mark Nelson
2015-04-17  0:38             ` Sage Weil
2015-04-17  0:47               ` Gregory Farnum
2015-04-17  0:53                 ` Sage Weil
2015-04-17  0:55                 ` Chen, Xiaoxi
2015-04-17  4:53               ` Haomai Wang
2015-04-17 15:28                 ` Sage Weil
2015-04-17 12:10               ` Mark Nelson
2015-04-17 14:08                 ` Chen, Xiaoxi
2015-04-17 14:20                   ` Haomai Wang
2015-04-17 14:29                     ` Chen, Xiaoxi
2015-04-17 14:34                       ` Mark Nelson
2015-04-17 14:40                 ` Chen, Xiaoxi
2015-04-17 15:25                   ` Mark Nelson
2015-04-17 16:05                     ` Sage Weil
2015-04-17 16:59                       ` Mark Nelson
2015-04-17 15:46                 ` Sage Weil
2015-04-18  3:34                   ` Mark Nelson [this message]
  -- strict thread matches above, loose matches on Subject: below --
2015-04-13 23:53 Somnath Roy
2015-04-14  0:06 ` Mark Nelson
2015-04-14  0:12   ` Somnath Roy
2015-04-14  0:21     ` Mark Nelson

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=5531D0B1.1070406@redhat.com \
    --to=mnelson@redhat.com \
    --cc=Somnath.Roy@sandisk.com \
    --cc=ceph-devel@vger.kernel.org \
    --cc=haomaiwang@gmail.com \
    --cc=sage@newdream.net \
    --cc=xiaoxi.chen@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 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.