Abstract

Grids as infrastructures offer access to computing, storage and other resources in a transparent way. The user does not have to be aware where and how the job is being executed. Grid clusters in particular are an interesting target for running computation-intensive calculations. Running MPI-parallel applications on such clusters is a logical approach that is of interest to both computer scientists and to engineers. This paper gives an overview of the issues connected to running MPI applications on a heterogenous Grid consisting of different clusters located at different sites within the Int.EU.Grid project. The role of a workload management system (WMS) for such a scenario, as well as important modifications that need to be made to a WMS oriented towards sequential batch jobs for better support of MPI applications and tools are discussed. In order to facilitate the adoption of MPI-parallel applications on heterogeneous Grids, the application developer should be made aware of performance problems, as well as MPI-standard issues within its code. Therefore tools for these issues are also supported within Int.EU.Grid. Also, the special case of running MPI applications on different clusters simultaneously as a more Grid-oriented computational approach is described.

Keywords

Cite this article

@article{cai-1825,
  author  = {Kiril Dichev and Sven Stork and Rainer Keller and Enol Fernández},
  title   = {MPI Support on the Grid},
  journal = {Computing and Informatics},
  volume  = {27},
  number  = {2},
  pages   = {213--222},
  year    = {2008},
  issn    = {1335-9150},
  url     = {https://cai.type.sk/content/2008/2/mpi-support-on-the-grid/}
}
Download .bib