From: Droebbel <droebbel.melta@gmx.de>
To: Tino Keitel <tino.keitel@gmx.de>
Cc: linux-kernel@vger.kernel.org
Subject: Re: [Problem] slow write to dvd-ram since 2.6.7-bk8
Date: Fri, 18 Feb 2005 23:21:10 +0100 [thread overview]
Message-ID: <1108765270.14370.22.camel@localhost.localdomain> (raw)
In-Reply-To: <20050217231048.GA4363@dose.home.local>
On Fr, 2005-02-18 at 00:10 +0100, Tino Keitel wrote:
>On Wed, Feb 16, 2005 at 23:29:24 +0100, Droebbel wrote:
>> On Mi, 2005-02-16 at 22:55 +0100, Droebbel wrote:
>>
>> >The vmscan-dont-reclaim-too-many-pages.patch led to the said reduction
>> >of writing speed. I reverse-applied it to 2.6.8.1, where it seems to
>> >solve the problem.
>>
>> Sorry, have to correct that: it seemed to help at my tests with dd
>> (write 1G of zeroes to a file). Copying a file with mc still shows
>> around 1.4MB/s. Could be worse, but is definitely not ok. It *is* better
>> with 2.6.7.
>
>Here are some numbers with my setup. I always wrote 1 GB of data to the
>same DVD-RAM disc (EMTEC), to the device directly and to a fresh ext2
>on
>the disc.
>
>kernel 2.6.10:
>
>$ time { sudo dd if=/dev/zero of=bigfile bs=64k count=16000 ; sync ; }
>
>real 32m5.025s
>
>$ time {sudo dd if=/dev/zero of=/dev/cdrom bs=64k count=16000 ; sync ;}
>
>real 29m41.980s
>
>kernel 2.6.7:
>
>$ time { sudo dd if=/dev/zero of=bigfile bs=64k count=16000 ; sync ; }
>
>real 13m23.688s
>
>$ time {sudo dd if=/dev/zero of=/dev/cdrom bs=64k count=16000 ; sync ;}
>
>real 13m14.609s
This is what I get:
2.6.8 to 2.6.10: about 30 min. I think that's clear now. I did not run
any mre test with that.
2.6.7 gives less than 10 minutes.
Reverse-Patched 2.6.8.1 and 2.6.10 about 9-11 min.
But what I think is interesting: Other than with 2.6.7, with my 2.6.10
the result seems to be highly dependent on other io activities. I came
to test that when I saw that writing to dvd-ram slowed down when reading
from a cdrom at the same time. System disk, cdrom and dvd-ram are
connected by buses as independent as possible: hda, hdd and scd0 via
on-chip ide2. hdc is inactive and spun down at the times of testing.
Some results (same command as yours, but writing to file only):
2.6.8.1 with vmscan-dont-reclaim-too-many-pages.patch
and vmscan-scan-sanity.patch reversed:
real 9m17.389s
real 10m11.271s
(run twice)
2.6.10, both patches reversed:
real 10m26.374s
same kernel, some io and high (but niced) system load by reading from
hda and gzipping into /dev/null
real 21m46.795s
same kernel, some io and low load by reading from cdrom (raw read with
dd as well) into /dev/null
real 22m11.639s
2.6.7 vanilla, some io and low load by reading from cdrom (raw read with
dd as well) into /dev/null
real 5m58.092
That's too fast - impossible on 3x media with verification. I'll check
that again. But it *really* seemed to be fast.
I also hat the impression that my tests with 2.6.10 and 2.6.8.1 were
much more promising when run from a rather basic testing system without
X, a bit closer to the 2.6.7 results. Haven't got time to check that
till monday. All the above results except the 2.6.7 are from Ubuntu
(Hoary) Systems with X and Gnome running.
Regards
David
prev parent reply other threads:[~2005-02-18 22:21 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-02-13 13:36 [Problem] slow write to dvd-ram since 2.6.7-bk8 Droebbel
2005-02-13 14:26 ` Wakko Warner
2005-02-14 8:53 ` Tino Keitel
2005-02-14 10:28 ` Droebbel
[not found] ` <1108376734.9495.8.camel@localhost.localdomain>
[not found] ` <20050214105332.GA7163@dose.home.local>
2005-02-14 12:34 ` Droebbel
[not found] ` <1108379351.9495.22.camel@localhost.localdomain>
[not found] ` <20050214111819.GA7691@dose.home.local>
2005-02-16 21:55 ` Droebbel
2005-02-16 22:29 ` Droebbel
2005-02-17 23:10 ` Tino Keitel
2005-02-18 22:21 ` Droebbel [this message]
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=1108765270.14370.22.camel@localhost.localdomain \
--to=droebbel.melta@gmx.de \
--cc=linux-kernel@vger.kernel.org \
--cc=tino.keitel@gmx.de \
/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).