Scalable multiscale modeling of platelets with 100 million particles

Changnian Han; Peng Zhang; Yicong Zhu; Guojing Cong; James R. Kozloski; Chih Chieh Yang; Leili Zhang; Yuefan Deng

doi:10.1007/s11227-022-04648-4

Scalable multiscale modeling of platelets with 100 million particles

Changnian Han, Peng Zhang, Yicong Zhu, Guojing Cong, James R. Kozloski, Chih Chieh Yang, Leili Zhang, Yuefan Deng

Learning Systems

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

3 Scopus citations

Abstract

We developed the core components of the AI-aided multiple time stepping algorithm for multiscale modeling of cell dynamics. This algorithm was implemented and analyzed on two supercomputer architectures with an application of simulating the aggregation of 250 platelets, or 102 million particles. To scale on these computers with complex memory and network architectures with GPUs, we devised a biomechanics-informed task mapping scheme to optimize load imbalance, communications, and memory utilization. Our simulations, scaling well up to 192 nodes on a Summit-like supercomputer with a peak speed of 11 petaflops, achieved a rate of 423 μ s/day which is 500 times faster than the conventional algorithm using static time step and this has enabled studies of record size blood clots at record spatial–temporal resolutions. Additionally, we discovered the sensitive dependence of the scalability and execution time on the methods of decomposition, CPU–GPU coupling, and task mapping.

Original language	English
Pages (from-to)	19707-19724
Number of pages	18
Journal	Journal of Supercomputing
Volume	78
Issue number	18
DOIs	https://doi.org/10.1007/s11227-022-04648-4
State	Published - Dec 2022

Funding

This project is supported by the SUNY-IBM Consortium Award, IPDyna: Intelligent Platelet Dynamics, FP00004096 (PI: Y. Deng). The simulations in this study were conducted on the AiMOS supercomputer at Rensselaer Polytechnic Institute and the SeaWulf Cluster at Stony Brook University (PIs: Y. Deng and P. Zhang). This project is supported by the SUNY-IBM Consortium Award, IPDyna: Intelligent Platelet Dynamics, FP00004096 (PI: Y. Deng). The simulations in this study were conducted on the AiMOS supercomputer at Rensselaer Polytechnic Institute and the SeaWulf Cluster at Stony Brook University (PIs: Y. Deng and P. Zhang).

Funders	Funder number
SUNY-IBM	FP00004096
Rensselaer Polytechnic Institute

Keywords

Artificial intelligence
Load balance
Multiscale modeling
Scalability

Access to Document

10.1007/s11227-022-04648-4

Cite this

@article{6f2b80824ce146e9b7c1281ec1424826,

title = "Scalable multiscale modeling of platelets with 100 million particles",

abstract = "We developed the core components of the AI-aided multiple time stepping algorithm for multiscale modeling of cell dynamics. This algorithm was implemented and analyzed on two supercomputer architectures with an application of simulating the aggregation of 250 platelets, or 102 million particles. To scale on these computers with complex memory and network architectures with GPUs, we devised a biomechanics-informed task mapping scheme to optimize load imbalance, communications, and memory utilization. Our simulations, scaling well up to 192 nodes on a Summit-like supercomputer with a peak speed of 11 petaflops, achieved a rate of 423 μ s/day which is 500 times faster than the conventional algorithm using static time step and this has enabled studies of record size blood clots at record spatial–temporal resolutions. Additionally, we discovered the sensitive dependence of the scalability and execution time on the methods of decomposition, CPU–GPU coupling, and task mapping.",

keywords = "Artificial intelligence, Load balance, Multiscale modeling, Scalability",

author = "Changnian Han and Peng Zhang and Yicong Zhu and Guojing Cong and Kozloski, {James R.} and Yang, {Chih Chieh} and Leili Zhang and Yuefan Deng",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.",

year = "2022",

month = dec,

doi = "10.1007/s11227-022-04648-4",

language = "English",

volume = "78",

pages = "19707--19724",

journal = "Journal of Supercomputing",

issn = "0920-8542",

publisher = "Springer",

number = "18",

}

TY - JOUR

T1 - Scalable multiscale modeling of platelets with 100 million particles

AU - Han, Changnian

AU - Zhang, Peng

AU - Zhu, Yicong

AU - Cong, Guojing

AU - Kozloski, James R.

AU - Yang, Chih Chieh

AU - Zhang, Leili

AU - Deng, Yuefan

PY - 2022/12

Y1 - 2022/12

N2 - We developed the core components of the AI-aided multiple time stepping algorithm for multiscale modeling of cell dynamics. This algorithm was implemented and analyzed on two supercomputer architectures with an application of simulating the aggregation of 250 platelets, or 102 million particles. To scale on these computers with complex memory and network architectures with GPUs, we devised a biomechanics-informed task mapping scheme to optimize load imbalance, communications, and memory utilization. Our simulations, scaling well up to 192 nodes on a Summit-like supercomputer with a peak speed of 11 petaflops, achieved a rate of 423 μ s/day which is 500 times faster than the conventional algorithm using static time step and this has enabled studies of record size blood clots at record spatial–temporal resolutions. Additionally, we discovered the sensitive dependence of the scalability and execution time on the methods of decomposition, CPU–GPU coupling, and task mapping.

AB - We developed the core components of the AI-aided multiple time stepping algorithm for multiscale modeling of cell dynamics. This algorithm was implemented and analyzed on two supercomputer architectures with an application of simulating the aggregation of 250 platelets, or 102 million particles. To scale on these computers with complex memory and network architectures with GPUs, we devised a biomechanics-informed task mapping scheme to optimize load imbalance, communications, and memory utilization. Our simulations, scaling well up to 192 nodes on a Summit-like supercomputer with a peak speed of 11 petaflops, achieved a rate of 423 μ s/day which is 500 times faster than the conventional algorithm using static time step and this has enabled studies of record size blood clots at record spatial–temporal resolutions. Additionally, we discovered the sensitive dependence of the scalability and execution time on the methods of decomposition, CPU–GPU coupling, and task mapping.

KW - Artificial intelligence

KW - Load balance

KW - Multiscale modeling

KW - Scalability

UR - http://www.scopus.com/inward/record.url?scp=85133183665&partnerID=8YFLogxK

U2 - 10.1007/s11227-022-04648-4

DO - 10.1007/s11227-022-04648-4

M3 - Article

AN - SCOPUS:85133183665

SN - 0920-8542

VL - 78

SP - 19707

EP - 19724

JO - Journal of Supercomputing

JF - Journal of Supercomputing

IS - 18

ER -

Scalable multiscale modeling of platelets with 100 million particles

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this