Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 15 Nov 2019 22:16:10 +0800
From:   Ming Lei <ming.lei@redhat.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     linux-block@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-xfs@vger.kernel.org, linux-kernel@vger.kernel.org,
        Jeff Moyer <jmoyer@redhat.com>,
        Dave Chinner <dchinner@redhat.com>,
        Eric Sandeen <sandeen@redhat.com>,
        Christoph Hellwig <hch@lst.de>, Jens Axboe <axboe@kernel.dk>,
        Ingo Molnar <mingo@redhat.com>, Tejun Heo <tj@kernel.org>
Subject: Re: single aio thread is migrated crazily by scheduler
Message-ID: <20191115141610.GA3283@ming.t460p>
References: <20191114113153.GB4213@ming.t460p>
 <20191114131434.GQ4114@hirez.programming.kicks-ass.net>
 <20191115000925.GB4847@ming.t460p>
MIME-Version: 1.0
In-Reply-To: <20191115000925.GB4847@ming.t460p>
User-Agent: Mutt/1.12.1 (2019-06-15)
Content-Type: text/plain; charset=WINDOWS-1252
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Nov 15, 2019 at 08:09:25AM +0800, Ming Lei wrote:
> On Thu, Nov 14, 2019 at 02:14:34PM +0100, Peter Zijlstra wrote:
> > On Thu, Nov 14, 2019 at 07:31:53PM +0800, Ming Lei wrote:
> > > Hi Guys,
> > >=20
> > > It is found that single AIO thread is migrated crazely by scheduler, =
and
> > > the migrate period can be < 10ms. Follows the test a):
> >=20
> > What does crazy mean? Does it cycle through the L3 mask?
> >=20
>=20
> The single thread AIO thread is migrated in several milliseconds once.

Today I found the migrate rate of single fio IO thread can reach
11~12K/sec when I run './xfs_complete 512' on another real machine
(single numa node, 8 cores).

And the number is very close to IOPS of the test, that said the fio
IO thread can be migrated once just when completing one IO on the
scsi_debug device.


Thanks,
Ming