mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Eric Dumazet <eric.dumazet@gmail.com>
To: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	linux-kernel@vger.kernel.org, aquini@redhat.com,
	lwoodman@redhat.com, jeremy@goop.org,
	Jan Beulich <JBeulich@novell.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Tom Herbert <therbert@google.com>
Subject: Re: [RFC PATCH 3/3 -v2] x86,smp: auto tune spinlock backoff delay factor
Date: Thu, 03 Jan 2013 10:17:48 -0800	[thread overview]
Message-ID: <1357237068.21409.25446.camel@edumazet-glaptop> (raw)
In-Reply-To: <CANN689HBx357M+7ge3SQ_xtnyiTKPY=1v0oR+DS9EBiak-2BQg@mail.gmail.com>

On Sat, 2012-12-29 at 02:27 -0800, Michel Lespinasse wrote:
> On Wed, Dec 26, 2012 at 11:10 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> > I did some tests with your patches with following configuration :
> >
> > tc qdisc add dev eth0 root htb r2q 1000 default 3
> > (to force a contention on qdisc lock, even with a multi queue net
> > device)
> >
> > and 24 concurrent "netperf -t UDP_STREAM -H other_machine -- -m 128"
> >
> > Machine : 2 Intel(R) Xeon(R) CPU X5660  @ 2.80GHz
> > (24 threads), and a fast NIC (10Gbps)
> >
> > Resulting in a 13 % regression (676 Mbits -> 595 Mbits)
> 
> I've been trying to use this workload on a similar machine. I am
> getting some confusing results however:
> 
> with 24 concurrent netperf -t UDP_STREAM -H $target -- -m 128 -R 1 , I
> am seeing some non-trivial run-to-run performance variation - about 5%
> in v3.7 baseline, but very significant after applying rik's 3 patches.
> my last few runs gave me results of 890.92, 1073.74, 963.13, 1234.41,
> 754.18, 893.82. This is generally better than what I'm getting with
> baseline, but the variance is huge (which is somewhat surprising given
> that rik's patches don't have the issue of hash collisions).

You mean that with Rik's patch, there is definitely an issue, as it has
a single bucket. Chances of collisions are high.


Your numbers being very random, I suspect you might hit another limit.

My tests involved a NIC with 24 transmit queues, to remove the per TX
queue lock out of the bench equation.

My guess is you use a NIC with 4 or 8 TX queues.

"ethtool -S eth0" would probably give some hints.

>  Also,
> this is significant in that I am not seeing the regression you were
> observing with just these 3 patches.
> 
> If I add a 1 second delay in the netperf command line (netperf -t
> UDP_STREAM -s 1 -H lpk18 -- -m 128 -R 1), I am seeing a very constant
> 660 Mbps result, but then I don't see any benefit from applying rik's
> patches. I have no explanation for these results, but I am getting
> them very consistently...
> 
> > In this workload we have at least two contended spinlocks, with
> > different delays. (spinlocks are not held for the same duration)
> 
> Just to confirm, I believe you are refering to qdisc->q.lock and
> qdisc->busylock ?
> 

Yes.




  reply	other threads:[~2013-01-03 18:23 UTC|newest]

Thread overview: 53+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-12-21 23:49 [RFC PATCH 0/3] x86,smp: make ticket spinlock proportional backoff w/ auto tuning Rik van Riel
2012-12-21 23:50 ` [RFC PATCH 1/3] x86,smp: move waiting on contended lock out of line Rik van Riel
2012-12-22  3:05   ` Steven Rostedt
2012-12-22  4:40   ` Michel Lespinasse
2012-12-22  4:48     ` Rik van Riel
2012-12-23 22:52   ` Rafael Aquini
2012-12-21 23:51 ` [RFC PATCH 2/3] x86,smp: proportional backoff for ticket spinlocks Rik van Riel
2012-12-22  3:07   ` Steven Rostedt
2012-12-22  3:14     ` Steven Rostedt
2012-12-22  3:47       ` Rik van Riel
2012-12-22  4:44   ` Michel Lespinasse
2012-12-23 22:55   ` Rafael Aquini
2012-12-21 23:51 ` [RFC PATCH 3/3] x86,smp: auto tune spinlock backoff delay factor Rik van Riel
2012-12-21 23:56   ` [RFC PATCH 3/3 -v2] " Rik van Riel
2012-12-22  0:18     ` Eric Dumazet
2012-12-22  2:43       ` Rik van Riel
2012-12-22  0:48     ` Eric Dumazet
2012-12-22  2:57       ` Rik van Riel
2012-12-22  3:29     ` Eric Dumazet
2012-12-22  3:44       ` Rik van Riel
2012-12-22  3:33     ` Steven Rostedt
2012-12-22  3:50       ` Rik van Riel
2012-12-26 19:10         ` Eric Dumazet
2012-12-26 19:27           ` Eric Dumazet
2012-12-26 19:51           ` Rik van Riel
2012-12-27  6:07             ` Michel Lespinasse
2012-12-27 14:27               ` Eric Dumazet
2012-12-27 14:35                 ` Rik van Riel
2012-12-27 18:41                   ` Jan Beulich
2012-12-27 19:09                     ` Rik van Riel
2013-01-03  9:05                       ` Jan Beulich
2013-01-03 13:24                         ` Steven Rostedt
2013-01-03 13:35                           ` Eric Dumazet
2013-01-03 15:32                             ` Steven Rostedt
2013-01-03 16:10                               ` Eric Dumazet
2013-01-03 16:45                                 ` Steven Rostedt
2013-01-03 17:54                                   ` Eric Dumazet
2012-12-27 18:49                   ` Eric Dumazet
2012-12-27 19:31                     ` Rik van Riel
2012-12-29  0:42                       ` Eric Dumazet
2012-12-29 10:27           ` Michel Lespinasse
2013-01-03 18:17             ` Eric Dumazet [this message]
2012-12-22  0:47   ` [RFC PATCH 3/3] " David Daney
2012-12-22  2:51     ` Rik van Riel
2012-12-22  3:49       ` Steven Rostedt
2012-12-22  3:58         ` Rik van Riel
2012-12-23 23:08           ` Rafael Aquini
2012-12-22  5:42   ` Michel Lespinasse
2012-12-22 14:32     ` Rik van Riel
2013-01-02  0:06 ` ticket spinlock proportional backoff experiments Michel Lespinasse
2013-01-02  0:09   ` [PATCH 1/2] x86,smp: simplify __ticket_spin_lock Michel Lespinasse
2013-01-02 15:31     ` Rik van Riel
2013-01-02  0:10   ` [PATCH 2/2] x86,smp: proportional backoff for ticket spinlocks Michel Lespinasse

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=1357237068.21409.25446.camel@edumazet-glaptop \
    --to=eric.dumazet@gmail.com \
    --cc=JBeulich@novell.com \
    --cc=aquini@redhat.com \
    --cc=jeremy@goop.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lwoodman@redhat.com \
    --cc=riel@redhat.com \
    --cc=rostedt@goodmis.org \
    --cc=tglx@linutronix.de \
    --cc=therbert@google.com \
    --cc=walken@google.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

all inboxes | Powered by JetHome®