Cluster Based Performance Enhancement of Plasma Simulation Codes

Abstract

Plasma Simulation codes are presently running on standalone systems in IPR. They are all compute intensive. Some highly compute intensive codes are running on super-computer. This is not cost effective in monetary terms. Also they need dedicated cluster which is available in super-computer. Thus there is a need to create HPC environment compatible to run them, here itself in IPR. It is difficult to have dedicated systems for such codes, so grid environment has to be created. All the codes need to be converted into cluster compatible code. Operating systems on each nodes in cluster may differ, like Ubuntu, fedora, red het etc. Thus configuration may differ. Also some code may use PVM and some may use MPI. MPI used by each code may also differ, vise Intel Mpi, LAM Mpi, HP Mpi etc. Hence an automatic cluster configuration tool is required which may serve the purpose. As the number of nodes in cluster configurations grows, installation, configuration, and administration becomes more challenging.