MaxSpeLL/dt-openusd-showcase-case03-airflow

Case 03 — Data Center Airflow Runtime Dataset

Overview

This dataset contains precomputed airflow velocity fields for Case 03 — Data Center, an OpenUSD / NVIDIA Omniverse digital-twin visualization project.

The velocity fields are generated in Houdini, resampled to a runtime-oriented grid, exported as temporal VTI assets, imported through NVIDIA Kit-CAE, and used by NVIDIA Flow to advect a passive smoke tracer inside the Digital Twin Runtime Suite (DTRS).

This repository contains runtime data, not the Houdini source simulation.

Current Dataset

Property	Value
Scope	server
Operating state	load_normal
Temporal samples	80
Source simulation rate	50 FPS
Export interval	every 10 source frames
Runtime velocity-field rate	5 Hz
Sample interval	0.2 s
Loop duration	16.0 s
Runtime grid	184 × 72 × 232
Field	vel
Format	VTI
Current published dataset size	~2.44 GB

---

Pipeline

Houdini airflow simulation
        ↓
Runtime-resampled velocity field
        ↓
Temporal VTI sequence
        ↓
NVIDIA Kit-CAE
        ↓
NVIDIA Flow DataSetEmitter
        ↓
Passive smoke tracer
        ↓
Digital Twin Runtime Suite (DTRS)

The precomputed Houdini velocity field remains the source of airflow motion.

NVIDIA Flow is used as the runtime visualization layer: smoke is injected as a passive tracer and advected by the imported velocity field.

---

Dataset Structure

The current published dataset is:

airflow_datasets/
└── 01_server/
    └── 02_load_normal/
        ├── manifest.toml
        ├── server_airflow_velocity_1001.vti
        ├── server_airflow_velocity_1011.vti
        ├── server_airflow_velocity_1021.vti
        └── ...

The numeric directory prefixes are only used for human-readable ordering.

DTRS does not identify datasets from folder names. It discovers manifest.toml files and resolves datasets from their semantic scope and state fields.

---

Manifest Contract

Each temporal airflow dataset contains a small manifest.toml describing its runtime contract.

The current dataset uses:

scope = "server"
state = "load_normal"

source_fps = 50
sample_step_frames = 10
sample_rate_hz = 5

sample_count = 80
grid = [184, 72, 232]

DTRS uses the manifest to:

• identify the requested scope and operating state;
• discover the VTI sequence automatically;
• validate sample count and frame spacing;
• derive runtime temporal cadence;
• validate the expected VTI grid;
• configure the temporal loop without hardcoded file lists.

For this dataset:

50 source frames / second
÷ 10-frame sampling interval
= 5 velocity-field updates / second

---

Temporal Sampling Trade-off

The original Houdini simulation is evaluated at 50 FPS, while the distributed runtime dataset samples the velocity field at 5 Hz.

This is an intentional runtime and distribution trade-off.

A denser temporal sequence is technically possible. For example, increasing the runtime field rate from 5 Hz to 10 Hz would approximately double the number of VTI assets required for the same simulation duration and grid resolution.

That cost becomes significant when the complete Case 03 dataset is expanded across multiple operating states and multiple simulation scopes:

server
    idle
    load_normal
    load_surge
    critical

rack
    idle
    load_normal
    load_surge
    critical

room
    idle
    load_normal
    load_surge
    critical

The current 5 Hz rate was selected as a practical compromise between:

• temporal smoothness;
• dataset size;
• transfer cost;
• reproducibility;
• runtime usability.

The higher-frequency Houdini source simulation remains available as the authoring source, so denser runtime exports can be generated when a specific validation or visualization task requires them.

---

Reproducing the Runtime Dataset

The related DTRS source repository is:

https://github.com/MSP014/dt-openusd-showcase-case03-dc

After cloning the project, the airflow dataset is expected under:

assets/_external/airflow_datasets/

Using the Hugging Face CLI from the root of the Case 03 repository:

hf download MaxSpeLL/dt-openusd-showcase-case03-airflow --repo-type dataset --local-dir assets/_external/airflow_datasets --include "01_server/**"

The resulting structure should contain:

assets/_external/
└── airflow_datasets/
    └── 01_server/
        └── 02_load_normal/
            ├── manifest.toml
            └── server_airflow_velocity_*.vti

DTRS discovers the dataset through the manifest; no manual list of VTI paths is required.

---

Current and Planned Coverage

Scope	Idle	Normal	Surge	Critical
Server	Planned	Available	Planned	Planned
Rack	Planned	Planned	Planned	Planned
Room	Planned	Planned	Planned	Planned

The current release therefore represents the first runtime airflow dataset for the project rather than the final complete operating-state library.

---

Intended Use

This dataset is intended for:

• reproducing the Case 03 DTRS airflow visualization;
• evaluating temporal engineering-field visualization through Kit-CAE and NVIDIA Flow;
• demonstrating a manifest-driven external simulation-data workflow;
• technical-art and digital-twin pipeline experimentation.

It is not presented as a general-purpose CFD validation benchmark.

The VTI sequence represents a precomputed velocity field prepared for interactive visualization. NVIDIA Flow provides the runtime tracer visualization rather than replacing the source airflow simulation.

---

Related Repository

Case 03 — Data Center / Digital Twin Runtime Suite

https://github.com/MSP014/dt-openusd-showcase-case03-dc

The GitHub repository contains the application code, OpenUSD/runtime integration, documentation, validation logic and reproducibility instructions.

Large temporal airflow assets are distributed separately through this Hugging Face dataset repository.

---

License

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Reuse and redistribution are permitted with appropriate attribution.