[mpich2-dev] ADIO layering

Rob Ross rross at mcs.anl.gov
Wed Dec 2 11:25:51 CST 2009


Hi John,

You're right that it would be useful to have a "stacking" concept in  
ROMIO. There is no such thing though, unfortunately.

You might find this paper helpful as a guide for this kind of work; it  
discusses some of the issues:
  
  
-------------- next part --------------
A non-text attachment was scrubbed...
Name: kimpe_logfs.pdf
Type: application/pdf
Size: 129524 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/mpich2-dev/attachments/20091202/e0711762/attachment-0001.pdf>
-------------- next part --------------


I've CC'd Dries, who did that work. He is the right person to answer  
these questions.

Regards,

Rob

On Dec 2, 2009, at 10:58 AM, John Bent wrote:

> Hey All,
>
> I'm coding up a PLFS ADIO layer and I'm realizing that one drawback is
> that codes can't leverage multiple ADIO layers.  Is this a concern?  I
> guess no-one needs PVFS/PLFS since PVFS doesn't suffer from an N-1
> problem.  And no-one needs PLFS/Lustre or PLFS/PanFS since both PanFS
> and Lustre have POSIX mount points.  I just thought about this and was
> wondering if there are other concerns that I haven't thought of.
>
> Also, which is probably the best skeletal ADIO layer to use as a
> template?
>
> Thanks,
>
> John



More information about the mpich2-dev mailing list