Short-term

Get baseline configuration running on production infrastructure
- Deploy on test sites (LAL, TCD, LIP, ...)
  - SGE (Imperial?)
  - LSF (?)
- Get feedback from users
- List of RPMs -- add to ETICS which will make it available automatically. ACTION: Cal
  - i2g-mpi-start
  - ...
- Recommend to SA3 for integration end of Jan 07
- Write up on GOC wiki
  - Recipe/hacks for sites (CEs) (with pointers to YAIM and Quattor impls.)
  - Page describing configuration and user environment (ACTION: Stephen)
- User software
  - JdlAndWrapper
  - Document on UIG site and replace MPI section in UG with link to this
Recommend change to RB to allow Normal jobs with >1 CPU (LCG and gLite)
- MPICH job type to be retained for compatibility but deprecated
- Look into compatibility of NodeNumber with other job types (e.g. checkpointable, interactive)

Next meeting

Allocation of CPUs and nodes (e.g. to ensure processes scheduled on the same node)
Changes to GLUE schema
- e.g. SWRTE per queue
Investigate whether having a jobtype on the RB with some MPI setup actually improves the user experience
- i.e. pre- and post-scripts would be all that the user has to write -- in gLite WMS?
Inter-site MPI -- EGEE to monitor I2G work -- may be installed on some EGEE sites even if not in production release
Investigate advertising and selecting compilers/libraries
Shared area that isn't a shared home