Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data Parallel Visualization and Rendering on the RAMSES Supercomputer with ANARI

Stefan Zellmann

TL;DR

This paper addresses rendering of distributed HPC data where a single node cannot hold the entire dataset. It analyzes data-parallel ray tracing and introduces DP-ANARI, a set of conventions that leverage ray queue cycling to render across MPI ranks without redistributing data. Through a RAMSES-based in-situ case study of the NASA Mars Lander dataset, integrated into OpenCOVER on 36 H100 GPUs, the work demonstrates interactive frame rates and highlights practical challenges such as uneven data distribution and cross-node occlusion. The findings show that ANARI can be effectively adapted to distributed HPC visualization, enabling scalable, interactive rendering on modern supercomputers and paving the way for VR-based in-situ visualization.

Abstract

3D visualization and rendering in HPC are very heterogenous applications, though fundamentally the tasks involved are well-defined and do not differ much from application to application. The Khronos Group's ANARI standard seeks to consolidate 3D rendering across sci-vis applications. This paper makes an effort to convey challenges of 3D rendering and visualization with ANARI in the context of HPC, where the data does not fit within a single node or GPU but must be distributed. It also provides a gentle introduction to parallel rendering concepts and challenges to practitioners from the field of HPC in general. Finally, we present a case study showcasing data parallel rendering on the new supercomputer RAMSES at the University of Cologne.

Data Parallel Visualization and Rendering on the RAMSES Supercomputer with ANARI

TL;DR

This paper addresses rendering of distributed HPC data where a single node cannot hold the entire dataset. It analyzes data-parallel ray tracing and introduces DP-ANARI, a set of conventions that leverage ray queue cycling to render across MPI ranks without redistributing data. Through a RAMSES-based in-situ case study of the NASA Mars Lander dataset, integrated into OpenCOVER on 36 H100 GPUs, the work demonstrates interactive frame rates and highlights practical challenges such as uneven data distribution and cross-node occlusion. The findings show that ANARI can be effectively adapted to distributed HPC visualization, enabling scalable, interactive rendering on modern supercomputers and paving the way for VR-based in-situ visualization.

Abstract

3D visualization and rendering in HPC are very heterogenous applications, though fundamentally the tasks involved are well-defined and do not differ much from application to application. The Khronos Group's ANARI standard seeks to consolidate 3D rendering across sci-vis applications. This paper makes an effort to convey challenges of 3D rendering and visualization with ANARI in the context of HPC, where the data does not fit within a single node or GPU but must be distributed. It also provides a gentle introduction to parallel rendering concepts and challenges to practitioners from the field of HPC in general. Finally, we present a case study showcasing data parallel rendering on the new supercomputer RAMSES at the University of Cologne.
Paper Structure (7 sections, 3 figures)

This paper contains 7 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Ray Tracing
  4. Data Parallel Ray Tracing
  5. Data Parallel Rendering with ANARI (DP-ANARI)
  6. Interactive In-Situ Visualization on RAMSES Using DP-ANARI
  7. Conclusion

Figures (3)

  • Figure 1: Scenario that presents a challenge for data parallel ray tracing. The data (tetraedra, finite elements, etc.) is distributed across multiple MPI ranks. Here, a ray was traced from the camera and intersects with the data on rank 0. A shadow ray is cast to determine if the intersection point requires shading. There is an occluder, but on rank!1, so that the secondary ray needs to be sent there first before the operation can complete.
  • Figure 2: Thin display client connecting to DP-ANARI distributed MPI renderer on Cologne's super computer RAMSES, here operating a display wall while RAMSES renders the NASA Mars Lander on 36 NVIDIA H100 GPUs.
  • Figure 3: Mars Lander data set with color coding indicating which finite elements go on which rank.