Parallel Program Simulator

Target Program - MPI, MPI-IO

Local Code

Communication Routines

Parallel I/O

One process per cnode

P0 P1 P2 P3

Target architecture

P0’ P1’

P2’ P3’

Simulator

Host architecture

MPI / MPI-IO

Simulation Threads

Parallel file system with dedicated ionodes

Standard UNIX file system

Previous slide

Next slide

Back to first slide

View graphic version