Newsgroups: comp.parallel.mpi,comp.parallel.pvm
From: dmb@bifrost.Lanl.GOV (David M. Beazley)
Subject: Thinking about parallel I/O
Organization: T-11, Los Alamos National Laboratory
Date: 10 Jul 1995 16:32:07 GMT
Message-ID: <DMB.95Jul10103207@bifrost.Lanl.GOV>


Having spent some time working on a message-passing application for the CM-5, T3D and
other machines, one of the biggest problems that has come up has been the
apparent nonstandardization of I/O on different machines.   My solution to this problem,
was to write my own parallel I/O wrapper library that implemented the equivalent of
CMMD parallel file I/O operations (since this is what we used in our first version of code on
the CM-5).  We then use this library on the CM-5, T3D, and workstations.  This has
eliminated alot of complexity and generally improved code portability.   In any case, this
has got me thinking about a few things....

1.   Is parallel I/O important to most users?   Since we tend to run large production
     simulations we have certainly found it to be useful.   It's also nice to have some
     support for user-interaction with a running program.

2.   What is the current state of parallel I/O libraries?  The last time I checked (which
     was about a year ago), PVM and MPI didn't seem to have any parallel I/O support.
     Has this changed recently?  Is there any I/O support in MPI-2?   (You'll have to excuse
     my ignorance here since I've mainly worked with CMMD and the Cray shared memory library.) 
     Does anyone have any pointers to recent work?
	
3.   Does anyone have any positive or negative experiences related to I/O using PVM
     and MPI that they can share?

Parallel I/O seems to be something that has generally been ignored so I'm mainly trying
to find out what people have done in this area.   I'm very curious to hear about how
others have solved this problem within the PVM/MPI environment.  If you'd like to share
your experience, please respond.  If there's interest, I'll post a summary of responses later.

Cheers,

Dave
--------------------------------------
Dave Beazley
Los Alamos National Laboratory
dmb@viking.lanl.gov









