% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@INPROCEEDINGS{Tashakor:1022046,
author = {Tashakor, Ghazal},
othercontributors = {Caviedes Voullieme, Daniel},
title = {{T}owards {E}xascale through {M}odularity {A}nd {I}/{O}
{M}anagement, {P}erformance study under {M}odular computing
with {TSMP}},
reportid = {FZJ-2024-01183},
year = {2024},
note = {Platform
link:https://www.hipeac.net/2024/munich/#/program/},
abstract = {This abstract highlights the pivotal role of the
Terrestrial Systems Modelling Platform (TSMP) as a core use
case in two pioneering projects, DEEP-SEA and IO-SEA, within
the context of the Modular Supercomputing Architecture
(MSA). The MSA, developed throughout the DEEP projects,
serves as a blueprint for heterogeneous HPC systems,
promoting the highest efficiency and scalability by
integrating different compute modules tailored to specific
performance characteristics for diverse workloads.The
Terrestrial Systems Modelling Platform (TSMP) stands as a
versatile, fully coupled Earth system model designed for
regional simulations, emphasizing complex interactions
within the geo-ecosystem. As an open community code, TSMP
integrates various atmospheric models like COSMO and ICON,
the Community Land Model (CLM), and the hydrological model
ParFlow. With modular coupling, TSMP supports multiple
programming languages, parallelization schemes, and hardware
architectures.In the DEEP-SEA project, TSMP is utilized as a
key use case to address the challenges of exascale computing
within the MSA. The modular approach of the MSA aligns with
TSMP's capabilities, allowing for the execution of different
components, such as CPU and GPU-enabled COSMO, CUDA-ported
ParFlow, and Fortran-based CLM, on the most suitable
platform. The focus within DEEP-SEA is on mapping these
specific components to optimize memory usage and
scalability. This approach enables TSMP to conduct
simulations at unprecedented resolutions and speeds,
leveraging the innovative capabilities of the DEEP
infrastructure.Simultaneously, in the IO-SEA project, TSMP
serves as a foundational use case to drive advancements in
data-centric I/O and workflow execution within the MSA. The
adaptability and scalability of the MSA align with the goals
of IO-SEA in optimizing data-centric workflows. The project
extends existing I/O instrumentation tools, such as Smart
burst buffering (SBB) and Data Access and Storage
application Interface (DASI), to accommodate the intricacies
of TSMP's coupled workflow.In conclusion, the integration of
TSMP within the DEEP projects exemplifies a collaborative
effort to address the challenges of exascale computing
within the innovative Modular Supercomputing Architecture.
TSMP's coupled approach aligns seamlessly with the
adaptability and scalability of the MSA, allowing for
enhanced efficiency and performance in the rapidly evolving
landscape of heterogeneous HPC systems. This collaborative
effort serves as a model for leveraging versatile Earth
system models to improve energy and resource use efficiency
in cutting-edge supercomputing infrastructures.},
month = {Jan},
date = {2024-01-16},
organization = {HiPEAC, Munich (Germany), 16 Jan 2024
- 19 Jan 2024},
subtyp = {Invited},
cin = {JSC / IBG-3 / IAS},
cid = {I:(DE-Juel1)JSC-20090406 / I:(DE-Juel1)IBG-3-20101118 /
I:(DE-Juel1)VDB1106},
pnm = {2A5 - Exascale Earth System Modeling (CARF - CCA)
(POF4-2A5) / 2173 - Agro-biogeosystems: controls, feedbacks
and impact (POF4-217)},
pid = {G:(DE-HGF)POF4-2A5 / G:(DE-HGF)POF4-2173},
experiment = {EXP:(DE-MLZ)SCG-20150203},
typ = {PUB:(DE-HGF)24},
doi = {10.34734/FZJ-2024-01183},
url = {https://juser.fz-juelich.de/record/1022046},
}