Workshop: Workshop on Operational Data Analytics in HPC (HPC-ODA)
Authors: Marketa Faltynkova and Petr Ptacek (IT4Innovations National Supercomputing Center; Technical University of Ostrava, Czech Republic); Kashaf Ad Dooja (University of Bologna); Milan Jaros and Ondrej Vysocky (IT4Innovations National Supercomputing Center; Technical University of Ostrava, Czech Republic); Josef Pasek (Brno University of Technology, IT4Innovations National Supercomputing Center); and Tomas Stratil, Petr Strakos, and Lubomir Riha (IT4Innovations National Supercomputing Center; Technical University of Ostrava, Czech Republic)
Abstract: Modern data centers generate large volumes of heterogeneous monitoring data. While conventional two-dimensional visualization tools are effective for many analytical tasks, they provide limited spatial context at system and facility scales, particularly for cooling and network infrastructures. This paper presents a framework for spatially contextualized visualization of monitoring data using a realistic 3D cluster model. The approach maps multi-layer monitoring data onto a 3D representation, enabling both system-level overviews and detailed node-level inspection. Metrics are encoded using color-based textures and chart-based visualizations. Blender serves as the visualization platform, providing extensibility, high-quality rendering, and support for interactive and immersive exploration. The framework is demonstrated using a realistic model of the Karolina cluster. A generalized pipeline for parametric model creation is also introduced. It transforms tabular input data into a structured JSON representation and subsequently into Universal Scene Description (USD) format, enabling reusable and extensible 3D model generation.
Back to Workshop on Operational Data Analytics in HPC (HPC-ODA) Archive Listing
Back to Full Workshop Archive Listing