Lustre discuss - Jul 2013 - read less than one stripe

Hi all,

If the I/O request is less than one stripe width, what will happen on the
OST?
Is the lustre gonna access the whole stripe?

Thanks,

Jialin Liu


_______________________________________________
Lustre-discuss mailing list
Lustre-discuss-aLEFhgZF4x6X6Mz3xDxJMA@public.gmane.org
http://lists.lustre.org/mailman/listinfo/lustre-discuss

On 2013-07-11, at 22:20, "Jaln"
<valiantljk-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
wrote:
> If the I/O request is less than one stripe width, what will happen on the
OST?
> Is the lustre gonna access the whole stripe?
The stripe size controls the amount of data written to one OST before moving to
the next OST. It doesn''t control the IO size on the OST itself. The
minimum IO size on the OST is 4096 bytes, the blocksize of the underlying
ldiskfs filesystem.

Cheers, Andreas

Thank you Andreas. Then how the blocks are distributed? Are they in a
round-robin fashion, supposing there are 6 disks.
Thanks,

Jialin


On Thu, Jul 11, 2013 at 10:06 PM, Dilger, Andreas
<andreas.dilger-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>wrote:
> On 2013-07-11, at 22:20, "Jaln"
<valiantljk-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> > If the I/O request is less than one stripe width, what will happen on
> the OST?
> > Is the lustre gonna access the whole stripe?
>
> The stripe size controls the amount of data written to one OST before
> moving to the next OST. It doesn''t control the IO size on the OST
itself.
> The minimum IO size on the OST is 4096 bytes, the blocksize of the
> underlying ldiskfs filesystem.
>
> Cheers, Andreas



-- 

Genius only means hard-working all one''s life


_______________________________________________
Lustre-discuss mailing list
Lustre-discuss-aLEFhgZF4x6X6Mz3xDxJMA@public.gmane.org
http://lists.lustre.org/mailman/listinfo/lustre-discuss