We provide a program in the “test" directory of the HDF5 1.10 distribution - 
twriteorder.c<https://svn.hdfgroup.org/hdf5/branches/hdf5_1_10_0/test/twriteorder.c>.
 See 
description<https://svn.hdfgroup.org/hdf5/branches/hdf5_1_10_0/test/SWMR_POSIX_Order_UG.txt>.

On NFS it fails pretty quickly. We couldn’t make it to fail on GPFS.  Please 
see the 
report<https://svn.hdfgroup.org/hdf5/branches/hdf5_1_10_0/test/POSIX_Order_Write_Test_Report.pdf>.

It would be interesting to run it on PVFS.

Elena
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Elena Pourmal  The HDF Group  http://hdfgroup.org
1800 So. Oak St., Suite 203, Champaign IL 61820
217.531.6112
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~




On Mar 29, 2016, at 11:22 AM, Rob Latham 
<[email protected]<mailto:[email protected]>> wrote:



On 03/16/2016 11:28 AM, Dana Robinson wrote:
Hi Andrew,

We do not currently test SWMR on Windows but we would expect it to work
on NTFS. Basically we need for write ordering to be preserved, which is
usually true for a straightforward local file system. Parallel file
systems like GPFS are also generally supported.

On the topic of parallel file systems, it is unclear if the PVFS (aka OrangeFS) 
file system is POSIX-enough to support SWMR.  PVFS does not implement full 
POSIX semantics, for performance reasons, but neither is it as sloppy as NFS is.

==rob

_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]<mailto:[email protected]>
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

Reply via email to