CaDS Seminar 2024 - Feb. 13
Dr. Ghazal Tashakor (SDL Terrestrial Systems)
Integration of TSMP as a Key Use Case in DEEP-SEA and IO-SEA Projects: Enhancing Scalability and Automation in Heterogeneous Environments within the Modular Supercomputing Architecture (MSA)
Abstract:
This abstract elucidates the significant role of the Terrestrial Systems Modelling Platform (TSMP) as a central use case in two pioneering initiatives, DEEP-SEA and IO-SEA, embedded within the framework of Modular Supercomputing Architecture (MSA). Developed within the DEEP projects, MSA serves as a blueprint for heterogeneous High-Performance Computing (HPC) systems, fostering optimal efficiency and scalability by integrating diverse compute modules tailored to specific performance characteristics for varied workloads.
TSMP emerges as a versatile, fully coupled Earth system model engineered for regional simulations, accentuating intricate interactions within the geo-ecosystem. As an open community code, TSMP integrates diverse atmospheric models such as COSMO and ICON, the Community Land Model (CLM), and the hydrological model ParFlow. With modular coupling, TSMP accommodates multiple programming languages, parallelization schemes, and hardware architectures.
Within the DEEP-SEA project, TSMP is harnessed as a pivotal use case to tackle the complexities of exascale computing within MSA. The modular ethos of MSA harmonizes with TSMP's capabilities, facilitating the execution of different components, including CPU and GPU-enabled COSMO, CUDA-ported ParFlow, and FORTRAN-based CLM, on the most suitable platform. The emphasis within DEEP-SEA lies in optimizing memory usage and scalability by mapping these specific components, enabling TSMP to conduct simulations at unparalleled resolutions and speeds, leveraging the innovative capacities of the DEEP infrastructure.
Concurrently, in the IO-SEA project, TSMP acts as a foundational use case propelling advancements in data-centric I/O and workflow execution within MSA. The adaptability and scalability of MSA align with the objectives of IO-SEA in refining data-centric workflows. The project extends existing I/O instrumentation tools like Smart burst buffering (SBB) and Data Access and Storage Application Interface (DASI) to accommodate the intricacies of TSMP's coupled workflow.
In summary, the integration of TSMP within the DEEP and IO SEA projects embodies a collaborative endeavor to confront the challenges of exascale computing within the innovative Modular Supercomputing Architecture. TSMP's coupled methodology seamlessly integrates with the adaptability and scalability of MSA, fostering heightened efficiency and performance in the ever-evolving landscape of heterogeneous HPC systems. This collaborative initiative serves as a paradigm for leveraging versatile Earth system models to enhance energy and resource utilization efficiency in state-of-the-art supercomputing infrastructures.