TY - CONF
T1 - InfiniBand performance review
T2 - 2004 USENIX Annual Technical Conference
AU - Benjegerdes, Troy R.
AU - Bode, Brett M.
N1 - This work was performed under auspices of the U. S. Department of Energy under contract W-7405-Eng-82 at Ames Laboratory operated by the Iowa State University of Science and Technology. Funding was provided by the Mathematical, Information and Computational Science division of the Office of Advanced Scientific Computing Research. We would also like to thank InfiniCon Systems for a hardware loan, Jeff Kirk at Mellanox for invaluable technical assistance, everyone contributing to the OpenIB.org project, and linux kernel developers who have taken an interest in InfiniBand.
This work was performed under auspices of the U. S. Department of Energy under contract W-7405-Eng-82 at Ames Laboratory operated by the Iowa State University of Science and Technology. Funding was provided by the Mathematical, Information and Computational Science division of the Office of Advanced Scientific Computing Research.
PY - 2004
Y1 - 2004
N2 - The InfiniBand interconnect has received a great deal of attention recently as a result of its use in the Virginia Tech cluster that placed 3rd on the Top500 list. However, InfiniBand hardware and software have been available for over a year and continue to evolve. In addition, InfiniBand is supported by a number of different vendors each of which is seeking to differentiate themselves in the marketplace with slightly different hardware and software offerings. This paper will examine many of the current hardware and software implementations, illustrating their similarities and differences. In addition we will demonstrate the sometimes dramatic effects of the various tuning parameters of the various software implementations.
AB - The InfiniBand interconnect has received a great deal of attention recently as a result of its use in the Virginia Tech cluster that placed 3rd on the Top500 list. However, InfiniBand hardware and software have been available for over a year and continue to evolve. In addition, InfiniBand is supported by a number of different vendors each of which is seeking to differentiate themselves in the marketplace with slightly different hardware and software offerings. This paper will examine many of the current hardware and software implementations, illustrating their similarities and differences. In addition we will demonstrate the sometimes dramatic effects of the various tuning parameters of the various software implementations.
UR - https://www.scopus.com/pages/publications/85091084668
UR - https://www.scopus.com/pages/publications/85091084668#tab=citedBy
M3 - Paper
AN - SCOPUS:85091084668
Y2 - 27 June 2004 through 2 July 2004
ER -