noop or deadline on an SSD?

noop or deadline on an SSD?

[Pedro Ribeiro]

[-- Attachment #1.1: Type: text/plain, Size: 632 bytes --]

Hi all,

I recently moved my system to an SSD.

I decided to do what the XFS FAQ says (
http://xfs.org/index.php/XFS_FAQ#Q:_I_want_to_tune_my_XFS_filesystems_for_.3Csomething.3E)
and I used all the defaults for creating the new XFS file systems.

However the paragraph above also reads: "As of kernel 3.2.12, the default
i/o scheduler, CFQ, will defeat much of the parallelization in XFS."

Ok, I also read that with an SSD I should be using noop or deadline. And
which one of these would perform better with XFS with an SSD? And while I'm
at it, which one would be best for a  mechanical drive?

Thanks in advance.

Regards,
Pedro

[-- Attachment #1.2: Type: text/html, Size: 1692 bytes --]

[-- Attachment #2: Type: text/plain, Size: 121 bytes --]

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: noop or deadline on an SSD?

[Stan Hoeppner]

On 4/21/2013 5:38 PM, Pedro Ribeiro wrote:

> Ok, I also read that with an SSD I should be using noop or deadline. And
> which one of these would perform better with XFS with an SSD? And while I'm
> at it, which one would be best for a  mechanical drive?

$ echo [noop|deadline] > /sys/block/sdX/queue/scheduler

Given the ease with which one can switch between the two, it's far
better to test each yourself with your hardware and workloads, than to
rely on any generic advice.

-- 
Stan

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs