Register

Speaker: Alex Razoumov (SFU)

Abstract: In this more advanced session, you will learn how to perform remote and large-scale visualization on the Alliance's HPC clusters, for cases where you do not want to -- or simply cannot -- download the dataset to your local computer.

Small-scale remote visualization can be accomplished using JupyterHub or Open OnDemand, running either standalone ParaView or a ParaView client within a remote desktop environment.

For larger-scale interactive visualization, we will run a parallel ParaView server for data processing and rendering on a cluster and connect to it from a ParaView client on your local computer. This setup enables interactive visualization of very large datasets (multi-GB and potentially TB-scale) without transferring data locally, and without being limited by your own system's memory or CPU/GPU resources.

We will explore workflows involving both time and camera animation. We will then save the interactive visualization pipeline as a script and run it remotely as a batch job on the cluster, performing all rendering off-screen (without opening any windows) to generate animation files.

Finally, we will discuss strategies for optimizing data for parallel I/O and the compute resources required to process very large datasets.

Prerequisites: To participate in the hands-on exercises, please install ParaView on your computer. You can download it from the official website. Familiarity with ParaView basics -- creating a visualization pipeline, applying filters, and scripting -- is assumed and covered in earlier sessions.