Conference Presentation (Other)

FZJ-2018-07702

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Parallel I/O: Benchmarking and common pitfalls

Lührs, S. (Corresponding author)FZJ* ; Ambrosino, F. (Contributor) ; Brzezniak, M. (Contributor) ; Frings, W. (Contributor)FZJ* ; Funel, A. (Contributor) ; Guarnieri, G. (Contributor) ; Haefele, M. (Contributor) ; Iannone, F. (Contributor) ; Paluszkiewicz, T. (Contributor) ; Sierocinski, K. (Contributor)

2018

Please use a persistent id in citations: http://hdl.handle.net/2128/21155

Abstract: Beside the computational scalability of an HPC application, its I/O behaviour can significantly influence the overall performance. This behaviour is influenced by many factors, e.g. by the implementation inside the application, the file system, the network and data hardware and software, as well as, since storage is usually a shared resource, other jobs running on the same cluster. Within EoCoE an I/O benchmarking activity was established to investigate how certain I/O patterns can influence the overall performance and how this behaviour can be reflected within the real world HPC applications to optimize the general I/O bandwidth. The talk will present the benchmarking methodology, common parallel I/O pitfalls, and it should guide HPC application developers in the context of parallel I/O problems.

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 511 - Computational Science and Mathematical Methods (POF3-511) (POF3-511)
2. EoCoE - Energy oriented Centre of Excellence for computer applications (676629) (676629)

Appears in the scientific report 2018

---

Database coverage:

---