Damaris: Leveraging Multicore Parallelism to Mask I/O Jitter

Matthieu Dorier 1 Gabriel Antoniu 2 Franck Cappello 3, 4, 5 Marc Snir 6 Leigh Orf 7
2 KerData - Scalable Storage for Clouds and Beyond
IRISA-D1 - SYSTÈMES LARGE ÉCHELLE, Inria Rennes – Bretagne Atlantique
3 GRAND-LARGE - Global parallel and distributed computing
CNRS - Centre National de la Recherche Scientifique : UMR8623, Inria Saclay - Ile de France, UP11 - Université Paris-Sud - Paris 11, LIFL - Laboratoire d'Informatique Fondamentale de Lille, LRI - Laboratoire de Recherche en Informatique
Abstract : With exascale computing on the horizon, the performance variability of I/O systems represents a key challenge in sustaining high performance. In many HPC applications, I/O is concurrently performed by all processes, which leads to I/O bursts. This causes resource contention and substantial variability of I/O performance, which significantly impacts the overall application performance together with the predictability of its run time. In this paper, we first illustrate the presence of I/O jitter on different platforms, and show the impact of different user-configurable parameters and I/O approaches on write performance variability. We then propose a new approach to I/O, called Damaris, which leverages dedicated I/O cores on each multicore SMP node, along with the use of shared-memory, to efficiently perform asynchronous data processing and I/O. We evaluate our approach on three different platforms including the Kraken Cray XT5 supercomputer, with the CM1 atmospheric model, which is one of the target HPC applications for the Blue Waters project. By overlapping I/O with computation and by gathering data into large files while avoiding synchronization between cores, our solution brings several benefits: 1) it fully hides the jitter as well as all I/O-related costs, which makes simulation performance predictable; 2) it increases the sustained write throughput by a factor of 15 compared to standard approaches; 3) it allows almost perfect scalability of the simulation where other I/O approaches fail to scale; 4) it enables a 600% compression ratio without any additional overhead, leading to a major reduction of storage requirements.
Type de document :
Rapport
[Research Report] RR-7706, INRIA. 2012, pp.36
Liste complète des métadonnées

Littérature citée [26 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/inria-00614597
Contributeur : Matthieu Dorier <>
Soumis le : lundi 9 avril 2012 - 14:29:07
Dernière modification le : vendredi 16 novembre 2018 - 01:39:18
Document(s) archivé(s) le : mardi 10 juillet 2012 - 02:20:20

Fichiers

RR-7706.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : inria-00614597, version 3

Citation

Matthieu Dorier, Gabriel Antoniu, Franck Cappello, Marc Snir, Leigh Orf. Damaris: Leveraging Multicore Parallelism to Mask I/O Jitter. [Research Report] RR-7706, INRIA. 2012, pp.36. 〈inria-00614597v3〉

Partager

Métriques

Consultations de la notice

1029

Téléchargements de fichiers

333