Disguise d3 RenderStream Cluster Rendering & Viewport Isolation for Multi-Node CAVE

General TouchDesigner Discussion
1

Hi everyone, I am currently architecting a large-scale CAVE project that integrates TouchDesigner into a Disguise d3 multi-node cluster environment using RenderStream. I am looking for community best practices regarding cluster rendering, viewport isolation, and workload distribution for this type of setup.

The Challenge:

In my testing, I’ve realized that unlike other engines like unreal, TouchDesigner does not natively slice the viewport or automatically distribute rendering across multiple nodes via RenderStream. To avoid rendering the full scene on every node—which leads to performance overhead and “doubled” outputs—I am looking for the recommended workflow to manually/auto handle frustum/viewpoint isolation.

Current Implementation & Questions:

I have successfully implemented Mesh Mapping with TD RenderStream on a single node, and the Camera COMP settings correctly align the scene origin and perspective. I am trying to determine how to scale this into a multi-node cluster:

  1. Mapping Strategy: For CAVE setups, is it possible to maintain Mesh Mapping in a multi-node RenderStream cluster, or is Feed/Direct Mapping per output the current industry standard? I am looking to avoid performance hits caused by full-scene rendering on every node while maintaining accuracy.
  2. Architecture: For a multi-node setup, is the preferred best practice to run a single TD project that dynamically handles per-node rendering (e.g., via logic based on the RenderStream Index/Channel), or is it more stable to manage separate project instances per node?
  3. Official Guidelines & Checklists: Are there any official TouchDesigner RenderStream optimization guides, “Setup Checklists,” or whitepapers available that cover multi-node performance tuning and resource management? I want to ensure my architecture follows recommended patterns for high-resolution, multi-node environments.

If anyone has experience with similar multi-node/CAVE workflows or can point me toward resources for RenderStream cluster best practices, I would appreciate your insight.Thank you in advance!