VisIVO

Fabio Vitello, Ugo Becciani, Claudio Gheller, Eva Sciacca (spoke1), Nicola Tuccari (spoke1) + TBD

Spoke 3 Technical Workshop, Trieste October 9 / 11, 2023

2005

2009

2010

2011

2013

2013

2014

2015

2017

2019

2021

VisIVO Desktop

VisIVO Server

VisIVO Web

VisIVO Science Gateway

VisIVO Library

Space Mission

VisIVO Mobile

Vialactea Visual

Analytic (VLVA)

VLVA - EOSC

Cirasa

VLVA - NEANIAS

Background: VisIVO

Scientific Rationale

• Next-generation facilities are expected to collect vast amounts of data that will enable astronomers to explore the Universe in unprecedented detail, extracting meaningful knowledge.

​

• In parallel, the advancement of astrophysics simulations has allowed scientists to create virtual representations of cosmic phenomena, further enhancing our understanding of the Universe.

​

VisIVO as KSP

VisIVO will be adapted to exploit HPC architectures for:

• in-situ visualization
• interactive visual analytics

In-Situ Visualization

Typically, the results of a simulation are visualized only after the simulation is complete and the results have been written to disk. This workflow is successful in many situations, but suffers the limitation of having separated simulation and visualization processes and incurs the overhead of lots of disk I/O.

Within an In-situ visualization service the simulation communicates more directly and simultaneously with the visualizer, providing the following benefits:

• Uses memory instead of disk to send data from a simulation to the visualizer, saving time by avoiding expensive disk operations.
• Can update the visualization data after each iteration of the simulation and easily generate visualization output for each iteration.
• Allows feedback from the visualization to control the simulation.

​

• VisIVO Server is an open source collection of visualization modules for fast rendering of 3D views of astrophysical datasets.
• Provide APIs allowing a code running on HPC systems to produce a set of images or movies directly using VisIVO with its internal data arrays without the need to produce intermediate files.

VisIVO Server

VisIVO as KSP: In-Situ Visualization

• Provide a framework for in-situ visualization allowing to visualize simulation data as it is generated, supporting various data formats including structured and unstructured data;

​

• Activities carried out with Spoke 1 are tailored to pursue the following objectives:
• Optimize and parallelize VisIVO Server modules to efficiently handle and process simulated data on Exascale computing resources,
• Enhance the portability of the VisIVO modular applications and their resource requirements,
• Take advantage of a more flexible resource exploitation over heterogeneous HPC facilities (including also mixed HPC-Cloud resources),
• Minimise data-movement overheads and improve I/O performances.

​

Spoke 1

In-Situ Visualization

• GADGET importer updated: parallelized for multi node/multi thread platforms using MPI and OpenMP.
• Filtering modules: start to investigate CUDA and OpenACC to exploit multi GPU platforms
• Viewer modules: to be optimized for emerging processor architectures, exploiting the VTK-m, a toolkit of scientific visualization algorithms for emerging processor architectures.
• Explore alternative solution for in-situ visualization e.g. Paraview Catalyst API

GADGET

Snapshot

Visual Analytics

Visualizing and interact with big-data products using a local visualization software on user’s machine becomes less efficient as the data size increase, i.e. with a response time of several seconds or minutes for each single basic interaction such as changing field of view, zooming, panning, etc.

​

​

An alternative approach is to adopt a client-server architecture to enable visualization where the servers focus on computing and rendering the data, while the clients handle the communication with the backend, sends commands and instructions to be executed and then visualize the results on the user’s machine.

Vialactea Visual Analytic

​

• Provides access to radio & infrared surveys of the Galactic Plane archived in the Knowledge Base (VLKB);
• Supporting visualization of 2D images and 3D velocity datacubes (vol. renderings, slices);
• Enabling visualization of catalogues such as compact sources and filaments;
• Integrated with source finding services;

​

VisIVO as KSP: Visual Analytics

• Provide a client/server Visual Analytic service allowing interactive visualization of 2D and 3D datasets stored on a remote server;

​

• Activities are tailored to pursue the following objectives:
• Parallelize reader and tools.
• Implement a client/server architecture on the Vialactea Visual Analytic software,
• Minimise data-movement overheads and improve I/O performances,

​

Visual Analytics

• Parallel implementation of Fits file reader (MPI + SMP)

Technical Objectives, Methodologies and Solutions

Display

Client

ssh�client

sshd

Server

batch node

Cluster

front end

tunnel

tunnel

User PC

Data

Visual Analytics

• First implementation of a Client/Server application
• Server component is deployable ad Docker/Singularity container
• Client available for Linux (debian, fedora) and macOS

​

Next steps

• Explore alternative solution for in-situ visualization
• Continue the porting of existing filters and visualizations modules to exploit GPU (and multi GPU) platforms
• Explore different Volumetric rendering algorithms (e.g. NVIDIA IndeX)

​

• Call for ideas: How VisIVO KSP can support other KSP?