Open MPI: Goals, Concept, and Design of a Next Generation MPI Implementation

Open MPI: Goals, Concept, and Design of a Next Generation MPI Implementation

| Edgar Gabriel, Graham E. Fagg, George Bosilca, Thara Angskun, Jack J. Dongarra, Jeffrey M. Squyres, Vishal Sahay, Prabhanjan Kambadur, Brian Barrett, Andrew Lumsdaine, Ralph H. Castain, David J. Daniel, Richard L. Graham, Timothy S. Woodall
The paper introduces Open MPI, a new production-quality MPI-2 implementation designed to address the challenges of large-scale parallel computing. Open MPI is centered around the MPI Component Architecture (MCA), which provides a flexible and modular framework for managing and composing components. The architecture supports multiple concurrent user threads, robust handling of process and network failures, and efficient support for various interconnects. The paper outlines the goals, design, and implementation of Open MPI, highlighting its unique features and performance results. The first full release of Open MPI is planned for the 2004 Supercomputing Conference, with an initial beta release supporting most functionalities and a subset of network device drivers.The paper introduces Open MPI, a new production-quality MPI-2 implementation designed to address the challenges of large-scale parallel computing. Open MPI is centered around the MPI Component Architecture (MCA), which provides a flexible and modular framework for managing and composing components. The architecture supports multiple concurrent user threads, robust handling of process and network failures, and efficient support for various interconnects. The paper outlines the goals, design, and implementation of Open MPI, highlighting its unique features and performance results. The first full release of Open MPI is planned for the 2004 Supercomputing Conference, with an initial beta release supporting most functionalities and a subset of network device drivers.
Reach us at info@study.space
[slides and audio] Open MPI%3A Goals%2C Concept%2C and Design of a Next Generation MPI Implementation