Smarter Data Made Even Faster – Real-time 2D/3D Graphics from LBL’s Perlmutter
- No file copies from HPC are needed for visualization, reducing I/O load on HPC system
- Interactive 2D & 3D visualization without 3rd party software, reducing costs and time to discovery
- Fast volume file and extract output are also supported, utilizing parallel I/O techniques directly from each HPC rank to greatly speed up writing VTK/FVUNS files to disk. Kombyne™ eliminates the bottlenecks preventing people from using full fidelity volume data from solvers like OpenFOAM.
Check out our 3 1/2 minute video including the live demo:
Android tablet screen save with new Kombyne 2.0 GUI:
The simulation of over 50M cells, used a realistic atmospheric boundary layer inflow condition. The computation used hundreds of cores on Lawrence Berkeley Labs Perlmutter HPC system, currently number 12 on the HPC Top 500. The unsteady simulation ran throughout the SC23 event, with each new timestep taking about three seconds to compute. The interactive visualization created cut plane extracts in situ for every timestep and Kombyne™ transported these to the web client for rendering and interaction.
Access to Perlmutter was provided under the SENSEI project (Kombyne™ is compatible with SENSEI), DOE subcontract 7608512.
Kombyne™ is a component of our IntelliTwin™ Suite, the enterprise open source software that delivers a Realizable Digital Thread for HPC. Other components are SpectreUQ, delivering HPC-scale Uncertainty Quantification for any solver or workflow, IntelliTwin™, an HPC-scale distributed data management tool embracing metadata and provenance ontologies and the Data Analytics Suite, enabling frequency and modal decomposition workflows.