Default boot/root filesystem
rwheeler at redhat.com
Wed Sep 25 17:17:18 UTC 2013
On 09/25/2013 12:48 PM, Michael Cronenworth wrote:
> Chris Murphy wrote:
> He provides no reliable testing method. I don't consider his results to be
> scientific or useful. One blogger will not be enough evidence.
You should - and can - easily test your workload with your hardware/software
stack to see if the options make a difference.
There are known and documented performance issues with S-ATA's non-queued
ATA_TRIM commands. Each command issued causes a drain of the pending IO's to the
device in question which can be bad for IO intensive workloads. Of course, if
you don't push your storage to the max (or if you don't delete a lot), this
might have zero impact on your particular set up :)
The T13 standards body is working on queued versions of ATA_TRIM that should help.
Other SSD's (non-S-ATA ones), software stacks that handle discard and
enterprise/SCSI arrays that do, all will have different sweet spots.
If you get a chance to test, do share your results!
More information about the devel