10,000 performance models per minute – scalability of the UG4 simulation framework

Andreas Vogel; Alexandru Calotoiu; Alexandre Strube; Sebastian Reiter; Arne Nägel; Felix Wolf; Gabriel Wittum

doi:10.1007/978-3-662-48096-0_40

10,000 performance models per minute – scalability of the UG4 simulation framework

Andreas Vogel^*, Alexandru Calotoiu, Alexandre Strube, Sebastian Reiter, Arne Nägel, Felix Wolf, Gabriel Wittum

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

10 Scopus citations

Abstract

Numerically addressing scientific questions such as simulating drug diffusion through the human stratum corneum is a challenging task requiring complex codes and plenty of computational resources. The UG4 framework is used for such simulations, and though empirical tests have shown good scalability so far, its sheer size precludes analytical modeling of the entire code. We have developed a process which combines the power of our automated performance modeling method and the workflow manager JUBE to create insightful models for entire UG4 simulations. Examining three typical use cases, we identified and resolved a previously unknown latent scalability bottleneck. In collaboration with the code developers, we validated the performance expectations in each of the use cases, creating over 10,000 models in less than a minute, a feat previously impossible without our automation techniques.

Original language	English (US)
Title of host publication	Euro-Par 2015
Subtitle of host publication	Parallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings
Editors	Jesper Larsson Traff, Sascha Hunold, Francesco Versaci
Publisher	Springer Verlag
Pages	519-531
Number of pages	13
ISBN (Print)	9783662480953
DOIs	https://doi.org/10.1007/978-3-662-48096-0_40
State	Published - 2015
Externally published	Yes
Event	21st International Conference on Parallel and Distributed Computing, Euro-Par 2015 - Vienna, Austria Duration: Aug 24 2015 → Aug 28 2015

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9233
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	21st International Conference on Parallel and Distributed Computing, Euro-Par 2015
Country/Territory	Austria
City	Vienna
Period	08/24/15 → 08/28/15

Bibliographical note

Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 2015.

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-662-48096-0_40

Cite this

Vogel, A., Calotoiu, A., Strube, A., Reiter, S., Nägel, A., Wolf, F., & Wittum, G. (2015). 10,000 performance models per minute – scalability of the UG4 simulation framework. In J. L. Traff, S. Hunold, & F. Versaci (Eds.), Euro-Par 2015: Parallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings (pp. 519-531). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9233). Springer Verlag. https://doi.org/10.1007/978-3-662-48096-0_40

Vogel, Andreas ; Calotoiu, Alexandru ; Strube, Alexandre et al. / 10,000 performance models per minute – scalability of the UG4 simulation framework. Euro-Par 2015: Parallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings. editor / Jesper Larsson Traff ; Sascha Hunold ; Francesco Versaci. Springer Verlag, 2015. pp. 519-531 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{da3ba8f130634b7991043651c84b1cac,

title = "10,000 performance models per minute – scalability of the UG4 simulation framework",

abstract = "Numerically addressing scientific questions such as simulating drug diffusion through the human stratum corneum is a challenging task requiring complex codes and plenty of computational resources. The UG4 framework is used for such simulations, and though empirical tests have shown good scalability so far, its sheer size precludes analytical modeling of the entire code. We have developed a process which combines the power of our automated performance modeling method and the workflow manager JUBE to create insightful models for entire UG4 simulations. Examining three typical use cases, we identified and resolved a previously unknown latent scalability bottleneck. In collaboration with the code developers, we validated the performance expectations in each of the use cases, creating over 10,000 models in less than a minute, a feat previously impossible without our automation techniques.",

author = "Andreas Vogel and Alexandru Calotoiu and Alexandre Strube and Sebastian Reiter and Arne N{\"a}gel and Felix Wolf and Gabriel Wittum",

note = "Publisher Copyright: {\textcopyright} Springer-Verlag Berlin Heidelberg 2015.; 21st International Conference on Parallel and Distributed Computing, Euro-Par 2015 ; Conference date: 24-08-2015 Through 28-08-2015",

year = "2015",

doi = "10.1007/978-3-662-48096-0_40",

language = "English (US)",

isbn = "9783662480953",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "519--531",

editor = "Traff, {Jesper Larsson} and Sascha Hunold and Francesco Versaci",

booktitle = "Euro-Par 2015",

address = "Germany",

}

Vogel, A, Calotoiu, A, Strube, A, Reiter, S, Nägel, A, Wolf, F & Wittum, G 2015, 10,000 performance models per minute – scalability of the UG4 simulation framework. in JL Traff, S Hunold & F Versaci (eds), Euro-Par 2015: Parallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9233, Springer Verlag, pp. 519-531, 21st International Conference on Parallel and Distributed Computing, Euro-Par 2015, Vienna, Austria, 08/24/15. https://doi.org/10.1007/978-3-662-48096-0_40

10,000 performance models per minute – scalability of the UG4 simulation framework. / Vogel, Andreas; Calotoiu, Alexandru; Strube, Alexandre et al.
Euro-Par 2015: Parallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings. ed. / Jesper Larsson Traff; Sascha Hunold; Francesco Versaci. Springer Verlag, 2015. p. 519-531 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9233).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - 10,000 performance models per minute – scalability of the UG4 simulation framework

AU - Vogel, Andreas

AU - Calotoiu, Alexandru

AU - Strube, Alexandre

AU - Reiter, Sebastian

AU - Nägel, Arne

AU - Wolf, Felix

AU - Wittum, Gabriel

PY - 2015

Y1 - 2015

N2 - Numerically addressing scientific questions such as simulating drug diffusion through the human stratum corneum is a challenging task requiring complex codes and plenty of computational resources. The UG4 framework is used for such simulations, and though empirical tests have shown good scalability so far, its sheer size precludes analytical modeling of the entire code. We have developed a process which combines the power of our automated performance modeling method and the workflow manager JUBE to create insightful models for entire UG4 simulations. Examining three typical use cases, we identified and resolved a previously unknown latent scalability bottleneck. In collaboration with the code developers, we validated the performance expectations in each of the use cases, creating over 10,000 models in less than a minute, a feat previously impossible without our automation techniques.

AB - Numerically addressing scientific questions such as simulating drug diffusion through the human stratum corneum is a challenging task requiring complex codes and plenty of computational resources. The UG4 framework is used for such simulations, and though empirical tests have shown good scalability so far, its sheer size precludes analytical modeling of the entire code. We have developed a process which combines the power of our automated performance modeling method and the workflow manager JUBE to create insightful models for entire UG4 simulations. Examining three typical use cases, we identified and resolved a previously unknown latent scalability bottleneck. In collaboration with the code developers, we validated the performance expectations in each of the use cases, creating over 10,000 models in less than a minute, a feat previously impossible without our automation techniques.

UR - http://www.scopus.com/inward/record.url?scp=84944097516&partnerID=8YFLogxK

U2 - 10.1007/978-3-662-48096-0_40

DO - 10.1007/978-3-662-48096-0_40

M3 - Conference contribution

AN - SCOPUS:84944097516

SN - 9783662480953

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 519

EP - 531

BT - Euro-Par 2015

A2 - Traff, Jesper Larsson

A2 - Hunold, Sascha

A2 - Versaci, Francesco

PB - Springer Verlag

T2 - 21st International Conference on Parallel and Distributed Computing, Euro-Par 2015

Y2 - 24 August 2015 through 28 August 2015

ER -

Vogel A, Calotoiu A, Strube A, Reiter S, Nägel A, Wolf F et al. 10,000 performance models per minute – scalability of the UG4 simulation framework. In Traff JL, Hunold S, Versaci F, editors, Euro-Par 2015: Parallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings. Springer Verlag. 2015. p. 519-531. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-662-48096-0_40

10,000 performance models per minute – scalability of the UG4 simulation framework

Abstract

Publication series

Other

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this