Trace-based performance analysis for the petascale simulation code FLASH

Heike Jagode; Andreas Knüpfer; Jack Dongarra; Matthias Jurenz; Matthias S. Müller; Wolfgang E. Nagel

doi:10.1177/1094342010387806

Trace-based performance analysis for the petascale simulation code FLASH

Heike Jagode, Andreas Knüpfer, Jack Dongarra, Matthias Jurenz, Matthias S. Müller, Wolfgang E. Nagel

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

Performance analysis of applications on modern high-end petascale systems is increasingly challenging due to the rising complexity and quantity of the computing units. This paper presents a performance-analysis study using the Vampir performance-analysis tool suite, which examines application behavior as well as the fundamental system properties. This study was carried out on the Jaguar system at Oak Ridge National Laboratory, the fastest computer on the November 2009 Top500 list. We analyzed the FLASH simulation code that is designed to be scaled with tens of thousands of CPU cores, which means that using existing performance-analysis tools is very complex. The study reveals two classes of performance problems that are relevant for very high CPU counts: MPI communication and scalable I/O. For both, solutions are presented and verified. Finally, the paper proposes improvements and extensions for event tracing tools in order to allow scalability of the tools towards higher degrees of parallelism.

Original language	English
Pages (from-to)	428-439
Number of pages	12
Journal	International Journal of High Performance Computing Applications
Volume	25
Issue number	4
DOIs	https://doi.org/10.1177/1094342010387806
State	Published - Nov 2011
Externally published	Yes

Funding

This research was sponsored by the Office of Mathematical, Information, and Computational Sciences of the Office of Science, US Department of Energy, under Contract No. DE-AC05-00OR22725 with UT-Battelle, LLC. This work used resources of the National Center for Computational Sciences at Oak Ridge National Laboratory, which is supported by the Office of Science of the Department of Energy under Contract DE-AC05-00OR22725. This resource was made available via the Performance Evaluation and Analysis Consortium End Station, a Department of Energy INCITE project.

Funders	Funder number
Office of Science of the Department of Energy
U.S. Department of Energy	DE-AC05-00OR22725
Office of Science

Keywords

Vampir
collective I/O
collective MPI operations
event tracing
libNBC

Access to Document

10.1177/1094342010387806

Cite this

@article{861130f5842b430b85d8bc905ee61af4,

title = "Trace-based performance analysis for the petascale simulation code FLASH",

abstract = "Performance analysis of applications on modern high-end petascale systems is increasingly challenging due to the rising complexity and quantity of the computing units. This paper presents a performance-analysis study using the Vampir performance-analysis tool suite, which examines application behavior as well as the fundamental system properties. This study was carried out on the Jaguar system at Oak Ridge National Laboratory, the fastest computer on the November 2009 Top500 list. We analyzed the FLASH simulation code that is designed to be scaled with tens of thousands of CPU cores, which means that using existing performance-analysis tools is very complex. The study reveals two classes of performance problems that are relevant for very high CPU counts: MPI communication and scalable I/O. For both, solutions are presented and verified. Finally, the paper proposes improvements and extensions for event tracing tools in order to allow scalability of the tools towards higher degrees of parallelism.",

keywords = "Vampir, collective I/O, collective MPI operations, event tracing, libNBC",

author = "Heike Jagode and Andreas Kn{\"u}pfer and Jack Dongarra and Matthias Jurenz and M{\"u}ller, {Matthias S.} and Nagel, {Wolfgang E.}",

year = "2011",

month = nov,

doi = "10.1177/1094342010387806",

language = "English",

volume = "25",

pages = "428--439",

journal = "International Journal of High Performance Computing Applications",

issn = "1094-3420",

publisher = "SAGE Publications",

number = "4",

}

TY - JOUR

T1 - Trace-based performance analysis for the petascale simulation code FLASH

AU - Jagode, Heike

AU - Knüpfer, Andreas

AU - Dongarra, Jack

AU - Jurenz, Matthias

AU - Müller, Matthias S.

AU - Nagel, Wolfgang E.

PY - 2011/11

Y1 - 2011/11

N2 - Performance analysis of applications on modern high-end petascale systems is increasingly challenging due to the rising complexity and quantity of the computing units. This paper presents a performance-analysis study using the Vampir performance-analysis tool suite, which examines application behavior as well as the fundamental system properties. This study was carried out on the Jaguar system at Oak Ridge National Laboratory, the fastest computer on the November 2009 Top500 list. We analyzed the FLASH simulation code that is designed to be scaled with tens of thousands of CPU cores, which means that using existing performance-analysis tools is very complex. The study reveals two classes of performance problems that are relevant for very high CPU counts: MPI communication and scalable I/O. For both, solutions are presented and verified. Finally, the paper proposes improvements and extensions for event tracing tools in order to allow scalability of the tools towards higher degrees of parallelism.

AB - Performance analysis of applications on modern high-end petascale systems is increasingly challenging due to the rising complexity and quantity of the computing units. This paper presents a performance-analysis study using the Vampir performance-analysis tool suite, which examines application behavior as well as the fundamental system properties. This study was carried out on the Jaguar system at Oak Ridge National Laboratory, the fastest computer on the November 2009 Top500 list. We analyzed the FLASH simulation code that is designed to be scaled with tens of thousands of CPU cores, which means that using existing performance-analysis tools is very complex. The study reveals two classes of performance problems that are relevant for very high CPU counts: MPI communication and scalable I/O. For both, solutions are presented and verified. Finally, the paper proposes improvements and extensions for event tracing tools in order to allow scalability of the tools towards higher degrees of parallelism.

KW - Vampir

KW - collective I/O

KW - collective MPI operations

KW - event tracing

KW - libNBC

UR - http://www.scopus.com/inward/record.url?scp=82955223396&partnerID=8YFLogxK

U2 - 10.1177/1094342010387806

DO - 10.1177/1094342010387806

M3 - Article

AN - SCOPUS:82955223396

SN - 1094-3420

VL - 25

SP - 428

EP - 439

JO - International Journal of High Performance Computing Applications

JF - International Journal of High Performance Computing Applications

IS - 4

ER -

Trace-based performance analysis for the petascale simulation code FLASH

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this