MPI Working Group

Europe/Zurich
CERN 28-R-014 (CERN)

CERN 28-R-014

CERN

For those at CERN 28-R-014 is booked.
Jeroen Engelberts (SARA)
Description

Phone Details (which changed at the last second)

Scope

To recommend a method for deploying MPI support that will work for both users and site administrators. This was the mandate for the previous MPI working group. The additional element added in this working group is : how to get the allocated cores to be all on the same physical machine, or packed into as few physical machines as possible. It should be understood to what extent this is a preference and what extent it's a strict requirement.

Recommendations should be relative to the MPI support currently distributed in gLite and should consider both the lcg-CE and CREAM.

Recommendations should incorporate also the case where the jobs are not in fact MPI (or even parallel). For example, a collection of jobs analyzing the same data which is copied locally could benefit from this system.

The TF should also attempt to understand why so few sites support MPI at present, and recommend how this situation could be improved.

TASK FORCE

The task force should consist of a representative from the user community, one or more MPI experts, and a few site representatives who understand the technical implications of the solution.

The TF should coordinate with Steve Traylen's WN working group on common issues.

Members

  • Ugo Becciani
  • Dennis van Dok
  • Fokke Dijkstra
  • Karolis Eigelis
  • Jeroen Engelberts
  • Salvatore Montforte
  • Mehdi Sheikhalahashi (maybe not)
  • Steve Traylen
  • John Ryan
  • Roberto Alfieri
  • Vangelis Koukis
  • Roberto Barbera
document
Slides
The agenda of this meeting is empty