Speaker
Description
HEPS is scheduled to complete construction and enter operation in 2026. During operation, HEPS will generate massive, heterogeneous, and highly time-sensitive experimental datasets. Experimental workflows such as tomography imaging, diffraction and scattering analysis, and spectroscopy processing require large-scale data reconstruction, complex software environments, and intensive interactive analysis, posing significant challenges to computational resource scheduling, data processing efficiency, and analysis environment delivery.
To address the diversity of experimental data types, heterogeneous analysis workflows, and dynamically evolving computational demands, this work presents a unified data-driven computing platform for HEPS. The platform is built on a cloud-native architecture integrating virtualization, containerization, and high-performance computing resources, forming a comprehensive scientific computing service framework for online monitoring, offline reconstruction, large-scale batch analysis, and interactive data analysis.
The platform provides unified resource management and elastic scheduling of heterogeneous computing resources, enabling efficient utilization of CPU, GPU, storage, and network resources. Standardized data access mechanisms and encapsulated analysis environments simplify software deployment and improve the reproducibility and automation of scientific workflows. In addition to browser-based access, the platform also provides standardized third-party interfaces and programmable APIs, allowing users and external systems to dynamically request, create, and manage customized data analysis environments and automatically execute analysis tasks.
During the HEPS commissioning and trial operation phase, the platform has been stably deployed and validated across multiple beamlines and associated data processing tasks. It has continuously supported users in accessing computing resources and performing remote experimental data analysis.
