> Date: Thu, 26 Aug 2004 08:38:11 +0200
> From: "Michael Gauckler" <michael.gauckler_at_[hidden]>
> Subject: LAM: MPI_File_{open, write, close}
> To: <lam_at_[hidden]>
>
> Hello,
>
> I am unsure whether I understood correctly the idea of MPI_File_{open,
> write, close}. Could please somebody confirm or correct me if I am
> wrong.
>
> 1)
>
> The main feature of MPI_File_{open, write, close} is to control access
> to a file, so that several processes can read/write
> efficiently from/to the same file.
Yes, that's right.
> 2)
>
> When reading from/writing to a file from multiple processes,
> MPI_File_{open, write, close} needs a file system (nfs other similar)
> which is mounted on all nodes involved in writing/reading. This means,
> that MPI_File_{open, write, close} does not do any
> transportation of the
> data from one node to the other (the transportation is
> implicit done by the mounted file system).
The ROMIO implementation of MPI-IO that LAM uses requires that there be a
common file system that all processes have access to. It is not a
requirement of the MPI-IO specification; it's up to the implementation.
> 3)
>
> If I write to the local file system the files written are
> valid and can
> be read again, but only by the process running on the node
> with the file system (to which I wrote the files).
If you are writing to the local file system, only the local process on the
machine can write to it and read from it.
If you want to use MPI-IO on a cluster, I would suggest you at least use a
common NFS-mounted file system. But it would be much better performance-wise
if you installed a real parallel file system such as PVFS, which is freely
available from www.pvfs.org.
Rajeev
|
