ParaView @ KAUST

To facilitate the use of ParaView at KAUST, we have developed an example repository (ParaView Vignettes) that explains and demonstrates how to run ParaView on your local machine, Ibex, and Shaheen. Please check out this repository for best practices and examples of how to connect to a remote computing resource to visualizate your data, or to generate scripts to make visualization easier and automatic.

ParaView Vignettes Repo

ParaView_Vignettes

This repository serves two primary purposes for High Performance Computing (HPC) visualization:

1. The Vignettes: A collection of self-contained examples demonstrating how to run ParaView scripts in batch mode on HPC resources.

2. Interactive Guide: Documentation on configuring local ParaView clients to connect to remote KAUST HPC clusters (Ibex and Shaheen III).

Table of Contents

• Part 1: Running the Vignettes (Batch Processing)

  • Generic / Local Setup

  • KAUST Ibex Setup

  • KAUST Shaheen III Setup

• Part 2: Interactive ParaView (Client-Server Mode)

  • Client Installation & Prerequisites

  • Connection Setup

  • KAUST Connection Guide (GUI Options)

• Part 3: HPC Resource Strategy (Performance Cheat Sheet)

• Appendix (Example Details & Reference)

---

Part 1: Running the Vignettes

Use this section if you want to run the provided example scripts (ex01, ex02, etc.) on a cluster. These examples are designed to run in Batch Mode (non-interactively).

Generic / Local Setup

Use this for your local machine or non-KAUST clusters.

1. Clone the Repository:

   git clone [https://github.com/jameskress/Visualization_Vignettes.git](https://github.com/jameskress/Visualization_Vignettes.git)
   cd Visualization_Vignettes/ParaView_Vignettes
   

2. Environment Setup:

   # Load ParaView module (system dependent)
   module load paraview
   # Or source the provided environment script
   source ../MODULES.sh
   

3. Run an Example: Copy the template script inside an example folder (e.g., ex01/ex01_template_runScript.sbat), customize it for your scheduler, and submit it.

KAUST Ibex Setup

1. Connect: ssh <user>@glogin.ibex.kaust.edu.sa

2. Clone:

   cd /ibex/scratch/<username>/
   git clone [https://github.com/jameskress/Visualization_Vignettes.git](https://github.com/jameskress/Visualization_Vignettes.git)
   cd Visualization_Vignettes/ParaView_Vignettes
   

3. Run:

   module load paraview
   sbatch ex01/ex01_ibex_runScript.sbat
   

KAUST Shaheen III Setup

1. Connect: ssh <user>@shaheen.hpc.kaust.edu.sa

2. Clone:

   cd /scratch/<username>/
   git clone [https://github.com/jameskress/Visualization_Vignettes.git](https://github.com/jameskress/Visualization_Vignettes.git)
   cd Visualization_Vignettes/ParaView_Vignettes
   

3. Configure & Run:

   # You MUST edit the script to add your Project Account (e.g., k01)
   vim ex01/ex01_shaheen_runScript.sbat
   # Change: #SBATCH --account=k##
   
   sbatch ex01/ex01_shaheen_runScript.sbat
   

IMPORTANT: GPU Access & The “Video” Group If you intend to use the ppn partition (GPU nodes) on Shaheen III for hardware-accelerated rendering, you must be a member of the video Unix group.

• How to check: Run the command groups in your terminal. If you see video, you have access.

• How to apply: Send an email to help@hpc.kaust.edu.sa requesting addition to the video group for visualization purposes.

---

Part 2: Interactive ParaView

Use this section if you want to use the ParaView GUI on your laptop to visualize data stored on the supercomputer (Client-Server mode).

Client Installation & Prerequisites

Before connecting, you must prepare your local machine (laptop/desktop).

1. Install ParaView:

   • Download the client from ParaView.org.

   • Crucial: Your local client version MUST match the version on the HPC system (check module avail paraview on the cluster).

2. OS-Specific Requirements:

   • macOS Users: You MUST install XQuartz (X11). ParaView requires X11 to display the authentication window and handle the connection tunnel.

   • Windows Users: You need a terminal client to handle authentication.

     • CRITICAL PITFALL: The standard Windows Command Prompt or PowerShell often fails to correctly handle the reverse connection tunnel required by ParaView (due to issues with the native OpenSSH implementation).

     • SOLUTION: You MUST install and use PuTTY to ensure the connection works reliably.

   • Linux: No additional software is typically required.

Connection Setup

1. Get Server Configs (.pvsc):

   • Ibex: Download ibex_server.pvsc

   • Shaheen: Download shaheen_server.pvsc

2. Load Configs:

   • Open ParaView → File → Connect...

   • Click Load Servers → Select the downloaded .pvsc file.

3. Connect:

   • Select the server (e.g., shaheen) from the list and click Connect.

KAUST Connection Guide (GUI Options)

When you click Connect, a dialog will appear asking for job settings. Use this guide to choose the right options.

For Shaheen III

Option	Setting	Description
Queue Name	workq	Recommended. Standard exclusive access node. Uses CPU rendering (Mesa).
	shared	Good for small jobs or quick checks.
	ppn	GPU Node. Use only for heavy volume rendering. Requires video group (see Part 1).
Tasks Per Node	192	For workq. Uses all CPU cores for processing.
	128	For ppn (GPU nodes have fewer CPU cores).
	16	For shared.

For Ibex

Option	Setting	Description
Node Group	cpu	Recommended. Uses software rendering (Mesa). Good for 95% of tasks.
	gpu	Uses hardware rendering. Only for massive geometry or volume rendering.
Tasks/Rank	1 to 4	Keep this low. Setting this high splits the RAM too many times and causes crashes.

---

Part 3: HPC Resource Strategy

Use this cheat sheet to determine the resources you need for your job (Interactive or Batch).

1. Rendering Backend: Mesa vs. EGL

• Mesa (Software Rendering):

  • Use for: Isosurfaces, Slices, Clips, and general analysis.

  • Why: It is faster and more stable for geometry-heavy workflows on modern CPUs.

  • Target: Shaheen workq or Ibex cpu.

• EGL (Hardware/GPU Rendering):

  • Use for: Volume Rendering (Fog/Clouds/Fire) or massive triangle counts (>50M).

  • Target: Shaheen ppn or Ibex gpu.

  • Note: Shaheen ppn requires video group membership.

2. Shaheen Configuration Strategy

Metric: Tasks = CPU Threads

Data Size	Queue	Nodes	Tasks Setting
< 16 GB	shared	1	16
16 GB - 350 GB	workq	1	192
> 350 GB	workq	2+	192

3. Ibex Configuration Strategy

Metric: Tasks = MPI Ranks

Goal	Queue	Tasks Setting	Note
Standard Vis	batch	4	Balanced CPU/RAM usage.
High RAM	batch	1	Gives 100% of node RAM to a single process.

---

Appendix

Repository Structure

ex##_name/
├── ex##_name.py                  # Main ParaView Python script
├── ex##_template_runScript.sbat  # Template batch script
├── README.md                     # Specific documentation
└── helper_scripts/               # Utilities

Example Details

1. ex00_pvQuery: Loading data and querying mesh statistics/metadata.

2. ex01_pvScreenshot: Basic rendering pipeline and saving images.

3. ex02_pvAnimation: Camera path animation.

4. ex03_pvIsosurfaceAnimation: Animating filter parameters (Isovalues).

5. ex04_pvStreamlineAnimation: Flow visualization and particle tracing.

6. ex05_pvMultiTimeStepFile: Handling time-series datasets.

7. ex06_pvLargeData: Optimization for massive datasets (ghost cells, parallel rendering).

pvbatch vs. pvpython

• pvpython: Serial. Runs on one core. Use for testing on login nodes.

• pvbatch: Parallel. Runs with MPI. Always use this for these examples.