Re: bad performance with SSD since kernel version 2.6.32

[Robert Hancock <hancockrwd@gmail.com>]

CCing Rafael, this should be added to the list of known regressions.

On Mon, Feb 22, 2010 at 7:18 AM, Benjamin S. <sbenni@gmx.de> wrote:
>
> On Sun, 21 Feb 2010 16:00:35 -0600
> Robert Hancock <hancockrwd@gmail.com> wrote:
>
>> Hmm.. Well, that's not it then.. I suspect a bisection is likely the
>> easiest route at this point..
>
> fb1e75389bd06fd5987e9cda1b4e0305c782f854 is the first bad commit
> commit fb1e75389bd06fd5987e9cda1b4e0305c782f854
> Author: Jens Axboe <jens.axboe@oracle.com>
> Date:   Thu Jul 30 08:18:24 2009 +0200
>
>    block: improve queue_should_plug() by looking at IO depths
>
>    Instead of just checking whether this device uses block layer
>    tagging, we can improve the detection by looking at the maximum
>    queue depth it has reached. If that crosses 4, then deem it a
>    queuing device.
>
>    This is important on high IOPS devices, since plugging hurts
>    the performance there (it can be as much as 10-15% of the sys
>    time).
>
>    Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
>
>
>
>
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=fb1e75389bd06fd5987e9cda1b4e0305c782f854
>
>
> Without that patch my SSD (Super Talent Ultradrive GX MLC 64GB)
> reaches about 200MB/s sequentiell read. After applying the patch it
> reaches only 70MB/s.
>
>
>
> Note: Added Jens to CC.
>
>