All of lore.kernel.org
 help / color / mirror / Atom feed
From: Bart Van Assche <Bart.VanAssche@sandisk.com>
To: Paolo Valente <paolo.valente@linaro.org>
Cc: "aherrmann@suse.com" <aherrmann@suse.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-block@vger.kernel.org" <linux-block@vger.kernel.org>,
	"axboe@kernel.dk" <axboe@kernel.dk>
Subject: Re: bfq-mq performance comparison to cfq
Date: Wed, 19 Apr 2017 05:01:36 +0000	[thread overview]
Message-ID: <CY1PR0401MB15362FAB3D841CDF2C421A2981180@CY1PR0401MB1536.namprd04.prod.outlook.com> (raw)
In-Reply-To: B7819549-81C3-4952-A31D-5E4A0732AB14@linaro.org

On 04/11/17 00:29, Paolo Valente wrote:=0A=
>=0A=
>> Il giorno 10 apr 2017, alle ore 17:15, Bart Van Assche <bart.vanassche@s=
andisk.com> ha scritto:=0A=
>>=0A=
>> On Mon, 2017-04-10 at 11:55 +0200, Paolo Valente wrote:=0A=
>>> That said, if you do always want maximum throughput, even at the=0A=
>>> expense of latency, then just switch off low-latency heuristics, i.e.,=
=0A=
>>> set low_latency to 0.  Depending on the device, setting slice_ilde to=
=0A=
>>> 0 may help a lot too (as well as with CFQ).  If the throughput is=0A=
>>> still low also after forcing BFQ to an only-throughput mode, then you=
=0A=
>>> hit some bug, and I'll have a little more work to do ...=0A=
>>=0A=
>> Has it been considered to make applications tell the I/O scheduler=0A=
>> whether to optimize for latency or for throughput? It shouldn't be that=
=0A=
>> hard for window managers and shells to figure out whether or not a new=
=0A=
>> application that is being started is interactive or not. This would=0A=
>> require a mechanism that allows applications to provide such information=
=0A=
>> to the I/O scheduler. Wouldn't that be a better approach than the I/O=0A=
>> scheduler trying to guess whether or not an application is an interactiv=
e=0A=
>> application?=0A=
>=0A=
> IMO that would be an (or maybe the) optimal solution, in terms of both=0A=
> throughput and latency.  We have even developed a prototype doing what=0A=
> you propose, for Android.  Unfortunately, I have not yet succeeded in=0A=
> getting support, to turn it into candidate production code, or to make=0A=
> a similar solution for lsb-compliant systems.=0A=
=0A=
Hello Paolo,=0A=
=0A=
What API was used by the Android application to tell the I/O scheduler =0A=
to optimize for latency? Do you think that it would be sufficient if the =
=0A=
application uses the ioprio_set() system call to set the I/O priority to =
=0A=
IOPRIO_CLASS_RT?=0A=
=0A=
Thanks,=0A=
=0A=
Bart.=0A=

WARNING: multiple messages have this Message-ID (diff)
From: Bart Van Assche <Bart.VanAssche@sandisk.com>
To: Paolo Valente <paolo.valente@linaro.org>
Cc: "aherrmann@suse.com" <aherrmann@suse.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-block@vger.kernel.org" <linux-block@vger.kernel.org>,
	"axboe@kernel.dk" <axboe@kernel.dk>
Subject: Re: bfq-mq performance comparison to cfq
Date: Wed, 19 Apr 2017 05:01:36 +0000	[thread overview]
Message-ID: <CY1PR0401MB15362FAB3D841CDF2C421A2981180@CY1PR0401MB1536.namprd04.prod.outlook.com> (raw)
In-Reply-To: B7819549-81C3-4952-A31D-5E4A0732AB14@linaro.org

On 04/11/17 00:29, Paolo Valente wrote:
>
>> Il giorno 10 apr 2017, alle ore 17:15, Bart Van Assche <bart.vanassche@sandisk.com> ha scritto:
>>
>> On Mon, 2017-04-10 at 11:55 +0200, Paolo Valente wrote:
>>> That said, if you do always want maximum throughput, even at the
>>> expense of latency, then just switch off low-latency heuristics, i.e.,
>>> set low_latency to 0.  Depending on the device, setting slice_ilde to
>>> 0 may help a lot too (as well as with CFQ).  If the throughput is
>>> still low also after forcing BFQ to an only-throughput mode, then you
>>> hit some bug, and I'll have a little more work to do ...
>>
>> Has it been considered to make applications tell the I/O scheduler
>> whether to optimize for latency or for throughput? It shouldn't be that
>> hard for window managers and shells to figure out whether or not a new
>> application that is being started is interactive or not. This would
>> require a mechanism that allows applications to provide such information
>> to the I/O scheduler. Wouldn't that be a better approach than the I/O
>> scheduler trying to guess whether or not an application is an interactive
>> application?
>
> IMO that would be an (or maybe the) optimal solution, in terms of both
> throughput and latency.  We have even developed a prototype doing what
> you propose, for Android.  Unfortunately, I have not yet succeeded in
> getting support, to turn it into candidate production code, or to make
> a similar solution for lsb-compliant systems.

Hello Paolo,

What API was used by the Android application to tell the I/O scheduler 
to optimize for latency? Do you think that it would be sufficient if the 
application uses the ioprio_set() system call to set the I/O priority to 
IOPRIO_CLASS_RT?

Thanks,

Bart.

  reply	other threads:[~2017-04-19  5:01 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-04-10  9:05 bfq-mq performance comparison to cfq Andreas Herrmann
2017-04-10  9:55 ` Paolo Valente
2017-04-10  9:55   ` Paolo Valente
2017-04-10 15:15   ` Bart Van Assche
2017-04-10 15:15     ` Bart Van Assche
2017-04-11  7:29     ` Paolo Valente
2017-04-11  7:29       ` Paolo Valente
2017-04-19  5:01       ` Bart Van Assche [this message]
2017-04-19  5:01         ` Bart Van Assche
2017-04-19  7:02         ` Paolo Valente
2017-04-19  7:02           ` Paolo Valente
2017-04-19 15:43           ` Bart Van Assche
2017-04-19 15:43             ` Bart Van Assche
2017-04-25  9:40           ` Juri Lelli
2017-04-26  8:18             ` Paolo Valente
2017-04-26  8:18               ` Paolo Valente
2017-04-26 22:12               ` Bart Van Assche
2017-04-26 22:12                 ` Bart Van Assche
2017-04-11  7:26   ` Paolo Valente
2017-04-11  7:26     ` Paolo Valente
2017-04-11 16:31   ` Andreas Herrmann

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=CY1PR0401MB15362FAB3D841CDF2C421A2981180@CY1PR0401MB1536.namprd04.prod.outlook.com \
    --to=bart.vanassche@sandisk.com \
    --cc=aherrmann@suse.com \
    --cc=axboe@kernel.dk \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=paolo.valente@linaro.org \
    /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.