vfs_glusterfs changes to get rid of mutex use.

Ira Cooper ira at samba.org
Tue Jan 20 06:22:09 MST 2015


I'm sure there is 1001 different optimizations. If you have some benchmarks
that show a specific issue, I'd find that investing. But in the overall
cost of io, I'd have a hard time believing that we can detect a pipe over
an eventfd.  In a general purpose implementation, I could see detecting
such things.

Thanks,

-Ira

On Tuesday, January 20, 2015, Volker Lendecke <Volker.Lendecke at sernet.de>
wrote:

> On Mon, Jan 19, 2015 at 11:17:08PM -0500, Ira Cooper wrote:
> > This patch removes all of the posix mutexes used in the previous
> GlusterFS
> > AIO patch in favor of using a standard POSIX pipe.  This also simplifies
> > the actual code, and makes our fork safety issues more apparent.
>
> This looks much better, thanks a lot!
>
> One thing just struck me: If a pipe is much more expensive
> than an eventfd, can't we use an eventfd as the indicator?
> Per eventfd we might have 64 outstanding requests, each bit
> representing one request?
>
> Volker
>
> --
> SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
> phone: +49-551-370000-0, fax: +49-551-370000-9
> AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
> http://www.sernet.de, mailto:kontakt at sernet.de <javascript:;>
>


More information about the samba-technical mailing list