Uploaded March 2026 | Updated September 2026, 3 days ago
Speaker:
Ophélie Renaud
Abstract:
The Square Kilometre Array (SKA) is set to become the world’s largest radio telescope, with data volumes and real-time processing requirements that push current HPC infrastructures to their limits. This talk introduces the challenges imposed by SKA in terms of throughput, latency, and physical constraints such as memory and energy. To address these issues, we leverage PREESM, a model-based rapid prototyping tool, to simulate and deploy radio astronomy pipelines on a range of heterogeneous architectures, including traditional HPC systems and, as part of exploratory work, emerging RISC-V platforms.
We present a complete workflow combining small-scale deployment on real system, such as the NenuFAR interferomete, and large-scale simulations to assess scalability and performance in SKA-like scenarios. This dual approach enables us to benchmark current implementations against SKA’s real-time requirements and to identify critical limitations, especially those linked to memory usage and energy consumption.
The session will conclude by opening the discussion on future directions to improve scalability, including potential dynamic techniques and architecture-aware optimizations. The objective is not only to meet SKA’s constraints but also to anticipate the broader challenges of next-generation scientific pipelines under tight physical and operational constraints.
Speaker:
Ophélie Renaud
Abstract:
The Square Kilometre Array (SKA) is set to become the world’s largest radio telescope, with data volumes and real-time processing requirements that push current HPC infrastructures to their limits. This talk introduces the challenges imposed by SKA in terms of throughput, latency, and physical constraints such as memory and energy. To address these issues, we leverage PREESM, a model-based rapid prototyping tool, to simulate and deploy radio astronomy pipelines on a range of heterogeneous architectures, including traditional HPC systems and, as part of exploratory work, emerging RISC-V platforms.
We present a complete workflow combining small-scale deployment on real system, such as the NenuFAR interferomete, and large-scale simulations to assess scalability and performance in SKA-like scenarios. This dual approach enables us to benchmark current implementations against SKA’s real-time requirements and to identify critical limitations, especially those linked to memory usage and energy consumption.
The session will conclude by opening the discussion on future directions to improve scalability, including potential dynamic techniques and architecture-aware optimizations. The objective is not only to meet SKA’s constraints but also to anticipate the broader challenges of next-generation scientific pipelines under tight physical and operational constraints.
