dm-devel.redhat.com archive mirror
 help / color / mirror / Atom feed
From: Jens Axboe <axboe@kernel.dk>
To: Mike Snitzer <snitzer@redhat.com>,
	Sagi Grimberg <sagig@dev.mellanox.co.il>
Cc: "keith.busch@intel.com" <keith.busch@intel.com>,
	"linux-nvme@lists.infradead.org" <linux-nvme@lists.infradead.org>,
	Christoph Hellwig <hch@infradead.org>,
	device-mapper development <dm-devel@redhat.com>,
	Bart Van Assche <bart.vanassche@sandisk.com>
Subject: Re: dm-multipath low performance with blk-mq
Date: Wed, 27 Jan 2016 10:51:04 -0700	[thread overview]
Message-ID: <56A90388.1030105@kernel.dk> (raw)
In-Reply-To: <20160127174828.GA31802@redhat.com>

On 01/27/2016 10:48 AM, Mike Snitzer wrote:
> On Wed, Jan 27 2016 at  6:14am -0500,
> Sagi Grimberg <sagig@dev.mellanox.co.il> wrote:
>
>>
>>>> I don't think this is going to help __multipath_map() without some
>>>> configuration changes.  Now that we're running on already merged
>>>> requests instead of bios, the m->repeat_count is almost always set to 1,
>>>> so we call the path_selector every time, which means that we'll always
>>>> need the write lock. Bumping up the number of IOs we send before calling
>>>> the path selector again will give this patch a change to do some good
>>>> here.
>>>>
>>>> To do that you need to set:
>>>>
>>>> 	rr_min_io_rq <something_bigger_than_one>
>>>>
>>>> in the defaults section of /etc/multipath.conf and then reload the
>>>> multipathd service.
>>>>
>>>> The patch should hopefully help in multipath_busy() regardless of the
>>>> the rr_min_io_rq setting.
>>>
>>> This patch, while generic, is meant to help the blk-mq case.  A blk-mq
>>> request_queue doesn't have an elevator so the requests will not have
>>> seen merging.
>>>
>>> But yes, implied in the patch is the requirement to increase
>>> m->repeat_count via multipathd's rr_min_io_rq (I'll backfill a proper
>>> header once it is tested).
>>
>> I'll test it once I get some spare time (hopefully soon...)
>
> OK thanks.
>
> BTW, I _cannot_ get null_blk to come even close to your reported 1500K+
> IOPs on 2 "fast" systems I have access to.  Which arguments are you
> loading the null_blk module with?
>
> I've been using:
> modprobe null_blk gb=4 bs=4096 nr_devices=1 queue_mode=2 submit_queues=12
>
> On my 1 system is a 12 core single socket, single NUMA node with 12G of
> memory, I can only get ~500K read IOPs and ~85K write IOPs.
>
> On another much larger system with 72 cores and 4 NUMA nodes with 128G
> of memory, I can only get ~310K read IOPs and ~175K write IOPs.

Look at the completion method (irqmode) and completion time 
(completion_nsec).

-- 
Jens Axboe

  reply	other threads:[~2016-01-27 17:51 UTC|newest]

Thread overview: 65+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <569CD4D6.2040908@dev.mellanox.co.il>
2016-01-19 10:37 ` dm-multipath low performance with blk-mq Sagi Grimberg
2016-01-19 22:45   ` Mike Snitzer
2016-01-25 21:40     ` Mike Snitzer
2016-01-25 23:37       ` Benjamin Marzinski
2016-01-26 13:29         ` Mike Snitzer
2016-01-26 14:01           ` Hannes Reinecke
2016-01-26 14:47             ` Mike Snitzer
2016-01-26 14:56               ` Christoph Hellwig
2016-01-26 15:27                 ` Mike Snitzer
2016-01-26 15:57             ` Benjamin Marzinski
2016-01-27 11:14           ` Sagi Grimberg
2016-01-27 17:48             ` Mike Snitzer
2016-01-27 17:51               ` Jens Axboe [this message]
2016-01-27 18:16                 ` Mike Snitzer
2016-01-27 18:26                   ` Jens Axboe
2016-01-27 19:14                     ` Mike Snitzer
2016-01-27 19:50                       ` Jens Axboe
2016-01-27 17:56               ` Sagi Grimberg
2016-01-27 18:42                 ` Mike Snitzer
2016-01-27 19:49                   ` Jens Axboe
2016-01-27 20:45                     ` Mike Snitzer
2016-01-29 23:35                 ` Mike Snitzer
2016-01-30  8:52                   ` Hannes Reinecke
2016-01-30 19:12                     ` Mike Snitzer
2016-02-01  6:46                       ` Hannes Reinecke
2016-02-03 18:04                         ` Mike Snitzer
2016-02-03 18:24                           ` Mike Snitzer
2016-02-03 19:22                             ` Mike Snitzer
2016-02-04  6:54                             ` Hannes Reinecke
2016-02-04 13:54                               ` Mike Snitzer
2016-02-04 13:58                                 ` Hannes Reinecke
2016-02-04 14:09                                   ` Mike Snitzer
2016-02-04 14:32                                     ` Hannes Reinecke
2016-02-04 14:44                                       ` Mike Snitzer
2016-02-05 15:13                                 ` [RFC PATCH] dm: fix excessive dm-mq context switching Mike Snitzer
2016-02-05 18:05                                   ` Mike Snitzer
2016-02-05 19:19                                     ` Mike Snitzer
2016-02-07 15:41                                       ` Sagi Grimberg
2016-02-07 16:07                                         ` Mike Snitzer
2016-02-07 16:42                                           ` Sagi Grimberg
2016-02-07 16:37                                         ` Bart Van Assche
2016-02-07 16:43                                           ` Sagi Grimberg
2016-02-07 16:53                                             ` Mike Snitzer
2016-02-07 16:54                                             ` Sagi Grimberg
2016-02-07 17:20                                               ` Mike Snitzer
2016-02-08 12:21                                                 ` Sagi Grimberg
2016-02-08 14:34                                                   ` Mike Snitzer
2016-02-09  7:50                                                 ` Hannes Reinecke
2016-02-09 14:55                                                   ` Mike Snitzer
2016-02-09 15:32                                                     ` Hannes Reinecke
2016-02-10  0:45                                                       ` Mike Snitzer
2016-02-11  1:50                                                         ` RCU-ified dm-mpath for testing/review Mike Snitzer
2016-02-11  3:35                                                           ` Mike Snitzer
2016-02-11 15:34                                                           ` Mike Snitzer
2016-02-12 15:18                                                             ` Hannes Reinecke
2016-02-12 15:26                                                               ` Mike Snitzer
2016-02-12 16:04                                                                 ` Hannes Reinecke
2016-02-12 18:00                                                                   ` Mike Snitzer
2016-02-15  6:47                                                                     ` Hannes Reinecke
2016-01-26  1:49       ` dm-multipath low performance with blk-mq Benjamin Marzinski
2016-01-26 16:03       ` Mike Snitzer
2016-01-26 16:44         ` Christoph Hellwig
2016-01-27  2:09           ` Mike Snitzer
2016-01-27 11:10             ` Sagi Grimberg
2016-01-26 21:40         ` Benjamin Marzinski

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=56A90388.1030105@kernel.dk \
    --to=axboe@kernel.dk \
    --cc=bart.vanassche@sandisk.com \
    --cc=dm-devel@redhat.com \
    --cc=hch@infradead.org \
    --cc=keith.busch@intel.com \
    --cc=linux-nvme@lists.infradead.org \
    --cc=sagig@dev.mellanox.co.il \
    --cc=snitzer@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 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).