Open Stack

On 16/09/15 06:54, Chris Friesen wrote:
> On 09/15/2015 01:16 AM, Clay Gerrard wrote:
>> Idk about a chestnut, but there's this:
>>
>> http://lists.openstack.org/pipermail/openstack-operators/2011-October/000297.html
>>
> 
> That advice isn't quite accurate.  It says "Enabling barriers
> effectively turns all writes into Write-Through operations, so the write
> goes straight to the disk platter and you get little performance benefit
> from the raid card (which hurts a lot in terms of lost iops)."
> 
> That's actually not necessarily true.  It's complicated (See
> http://marc.info/?l=linux-scsi&m=136674609728515&w=2 for discussion) but
> from what I understand most hardware RAID controller vendors will not
> actually flush their non-volatile cache on a SCSI "SYNCHRONIZE CACHE"
> command.  They will ensure that the data is in the NV cache, and then
> return to the host.  (Which should be quick.)
> 
> This gets you the best of both worlds...you get the reliability of
> filesystem barriers enforcing ordering through the OS stack, with the
> performance benefit of not having to flush the data all the way down to
> the disks.
> 

Funny - the Ceph guys are discussing this very issue (in particular with
Intel SSD, but the thread is generally informative...and quite long):

https://www.mail-archive.com/ceph-users@lists.ceph.com/msg23037.html