Lustre discuss - Apr 2007 - Max obtainable performance for a single OST - THANKS!

On 4/14/07, Andreas Dilger adilger@clusterfs.com,
                  Nathaniel Rutman nathan@clusterfs.com wrote:

      > Use "lfs setstripe 0 0 3" for directory for ost[0,1,2],
"lfs
setstripe 0 3 3"
      > for directory for ost[3,4,5], etc.
         > You can use ''lfs setstripe'' to designate a
starting OST index,
for a
      > file or a directory.  With that, you could set up different
single-OSS
      > directories.

Thanks, this helps! In this way any file will indeed be confined to exactly
one OSS
while will be able to make use the combined power of three internal OSTs.

      > If you have more than a single client reading from an OSS at the
same
      > time, you don''t even need to do that -- use three OSTs per
OSS, with
      > everything single-striped.  Your multiple clients can use the
different
      > OSTs simultaneously, increasing your total performance.

True, but in this way the peak performance per file will be that of only one
OST,
which is only 300+ MB/sec in our case. Whereas with 3 OSTs striped we may
be able to obtain between 750 and 900+ MB/sec per file.

Thanks again for clarifying this issue - Andrei.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://mail.clusterfs.com/pipermail/lustre-discuss/attachments/20070414/4f2dc1d3/attachment.html

On Sat, 2007-14-04 at 12:51 +0200, Andrei Maslennikov
wrote:
> 
>  
> True, but in this way the peak performance per file will be that of
> only one OST,
> which is only 300+ MB/sec in our case. Whereas with 3 OSTs striped we
> may 
> be able to obtain between 750 and 900+ MB/sec per file.
Correct.  In a typical lustre installation though there are lots of
clients, so many clients contend for total OSS bandwidth anyway.  Are we
to understand that you will have only one client per lustre filesystem?

b.