Open Stack

On Fri, 2016-01-22 at 21:49 +0100, Premysl Kouril wrote:
> On 22 Jan 2016 17:43, "James Bottomley" <
> James.Bottomley at hansenpartnership.com> wrote:
> > The 3x difference in the benchmarks would seem to indicate a local
> > tuning or configuration problem, because it's not what most people 
> > see.  What the current benchmarks seem to show is about a 1-5%
> > difference between the directio and the direct to block paths 
> > depending on fstype, how its tuned, ioscheduler and underlying
> > device.
> 
> I will try to find problem in our configuration and re-run our 
> benchmarks. By chance do you have some more information (and possibly
> configuration) about the bechmarks you are mentioning?

Well, what I'd actually do (and don't say I put you up to this) is send
an email to linux-fsdevel at vger.kernel.org alleging a 60% performance
drop on O_DIRECT under whatever filesystem it is, along with the
figures you publish.  I bet this will get a swift response from the
lead developer for that filesystem and a demand to help you tune your
system better ...

James