Hello,
> 1. What is the main difference between MPICH2 and OpenMPI, and which is
> your (end-users') preference?
There are users which need both MPI-1 and MPI-2; they use both the MPICH
and OpenMPI flavors of both of those APIs.
> 3. Would it be acceptable (to your end-users) if the grid would only
> facilitate MPI with non-shared home? (i.e. due to scalability and
> reliability issues)
No, this is not acceptable. The users principally want a shared home
area between the MPI 'workers'. Some are willing to work with
non-shared home areas so long as this is transparent.
> 4. Would it be acceptable (to your end-users) if the grid would only
> facilitate one or two specific implementations of MPI? And what's the state
> of MPI-1.1 vs. MPI-2?
Both MPI-1 and MPI-2 must be supported. It is possible to limit this to
one implementation of both, but would be better to support a couple of
flavors.
> 5. What's your opinion on having a user compile his MPI program at the
> UI, versus having the WNs compile it?
Sending pre-compiled binaries is doomed to failure. The executables
tend to be quite sensitive to how they were compiled and with which
options. Sending binaries may work in limited cases, but it must be
possible to compile the code at the WN.
> 6. Should we (and how?) anticipate MS-MPI on Microsoft's pending HPC
> server products?
Looks like this only runs on Windows machines. This is unlikely to be a
priority until all of the EGEE client software runs under Windows.
Cal
|