CERN Telephone Conference
Title: MPI Working Group
Phone: +41 22 767 7000
Organizer: Charles Loomis
Time: 14:00 CERN
minutes
14:00
→
14:15
Organizational Issues15m
Mandate of the working group. Time scales for releases.
14:15
→
14:30
User Requirements15m
What do the users require in the short term and long term? Do users need MPICH-1? MPICH-2? Another MPI implementation? Shared home areas for input/output? Executable distribution to workers sufficient? Need for pre/post processing in same job (e.g. compiling the executable locally)? Need for sequence of MPI executables in the same job with same nodes? Multisite MPI? Particular compilation environment?
14:30
→
14:45
Site Requirements15m
Need to provide customized software environment (e.g. mpiexec for pbs, or for special hardware)? Limits on network connectivity?
What sites are willing to do tests? What batch systems, job managers, special hardware can we test?
14:45
→
15:00
Roadmap15m
What things can we likely get into the next gLite releases? What are the longer-term efforts? Who is available to write test cases? To modify configurations for systematic tests?