[MPI3-IO] [Mpi3-ft] Fault Tolerance & I/O Discussion

George Bosilca bosilca at eecs.utk.edu
Thu Feb 23 16:42:03 CST 2012


Due to the member-only policy on the I/O list, my previous attempt to send the new FT proposal bounced. This is a re-sent of the same message now that I subscribed.

Thanks,
  george.


> Based on the discussion and opinions expressed during this call, the new FT proposal has been extended to include wording about the I/O. The model for I/O is very synchronous, providing a single function related to I/O. This function can be used to invalidate a file, preempting any subsequent operation on it. The file will then have to be closed, and is necessary reopened based on a new communicator (a smaller version of the original one).
> 
> The current version if the FT proposal is attached to this email, and can be accessed in the working group wiki @ https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/User_Level_Failure_Mitigation.
> 
> We expect to have a final version of the proposal by Monday noon EST, as requested by the MPI_Forum to be able to have a first reading at the next meeting. We are looking forward to your comments.
> 
>  george.




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-io/attachments/20120223/c2b781f1/attachment-0002.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mpi3ft.pdf
Type: application/pdf
Size: 156070 bytes
Desc: not available
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-io/attachments/20120223/c2b781f1/attachment-0001.pdf>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-io/attachments/20120223/c2b781f1/attachment-0003.html>


More information about the mpiwg-io mailing list