Re: unfair io behaviour for high load interactive use still present in 2.6.31

From: Tobias Oetiker <tobi@oetiker.ch>
To: Corrado Zoccolo <czoccolo@gmail.com>
Cc: linux-kernel@vger.kernel.org
Subject: Re: unfair io behaviour for high load interactive use still present in 2.6.31
Date: Wed, 16 Sep 2009 14:40:36 +0200 (CEST)	[thread overview]
Message-ID: <alpine.DEB.2.00.0909161208100.14727@sebohet.brgvxre.pu> (raw)
In-Reply-To: <4e5e476b0909160029l296efe2fk7b873ac68f6bb6a3@mail.gmail.com>

[-- Attachment #1: Type: TEXT/PLAIN, Size: 2887 bytes --]

Hi Corrado,

Today Corrado Zoccolo wrote:

> Hi Tobias,
> On Tue, Sep 15, 2009 at 11:07 PM, Tobias Oetiker <tobi@oetiker.ch> wrote:
> > Today Corrado Zoccolo wrote:
> >
> >> On Tue, Sep 15, 2009 at 9:30 AM, Tobias Oetiker <tobi@oetiker.ch> wrote:
> >> > Below is an excerpt from iostat while the test is in full swing:
> >> >
> >> > * 2.6.31 (8 cpu x86_64, 24 GB Ram)
> >> > * scheduler = cfq
> >> > * iostat -m -x dm-5 5
> >> > * running in parallel on 3 lvm logical volumes
> >> >  on a single physical volume
> >> >
> >> > Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
> >> > ---------------------------------------------------------------------------------------------------------
> * with i/o controller patches, what happens if you put readers in one
> domain and writers in the other?

With the io-conroller and readers and writers in two groups, the
rMB/s and wMB/s come out roughly the same. there are two problems unfortunately

a) the interactive performance is still not good since directory access still seems to get.

b) processes normally do both, read AND write, so the distinction would have to be made at the io-controller level.

Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
---------------------------------------------------------------------------------------------------------
dm-5              0.00     0.00  549.80    0.00     2.15     0.00     8.00   577.75    6.82   1.82 100.00
dm-5              0.00     0.00  191.20 1333.60     0.75     5.21     8.00  1077.45  953.15   0.66 100.00
dm-5              0.00     0.00  457.20 1144.00     1.79     4.47     8.00  1820.56  836.86   0.62 100.00
dm-5              0.00     0.00  453.00 1328.80     1.77     5.19     8.00  1326.74 1168.99   0.56 100.00
dm-5              0.00     0.00  353.09  353.49     1.38     1.38     8.00  1405.80 1993.23   1.41  99.80
dm-5              0.00     0.00  280.40 1296.80     1.10     5.07     8.00  1248.95  747.99   0.63 100.00
dm-5              0.00     0.00  307.00 4557.20     1.20    17.80     8.00   985.92  211.25   0.21 100.00
dm-5              0.00     0.00    6.00    0.00     0.02     0.00     8.00   450.42 19116.00 166.67 100.00
dm-5              0.00     0.00  154.00  624.20     0.60     2.44     8.00   926.52 1119.77   1.29 100.00

cheers
tobi

> Are you willing to test some patches? I'm working on patches to reduce
> read latency, that may be interesting to you.
>
> Corrado
>
> > cheers
> > tobi
> >
> > --
> > Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
> > http://it.oetiker.ch tobi@oetiker.ch ++41 62 775 9902 / sb: -9900
>
>
>
>

-- 
Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
http://it.oetiker.ch tobi@oetiker.ch ++41 62 775 9902 / sb: -9900