| |
|
Title: |
VISS: Visualization Infrastructures for Scalable Scientific Computing |
|
Address: |
Herr Prof. Renato Pajarola
Institut für Informatik
Universität Zürich
Binzmühlestrasse 14
CH-8050 Zürich |
|
|
Project Duration: |
4/1/2007 - 3/31/2010 |
|
Amount Granted: |
CHF 160,550.00 |
|
|
Funding Instrument: |
Project Support: Independent Basic Research
|
Principal Applicant
Pajarola Renato
Zürich
» Details
|
Co-applicant(s)
Baldridge Kim K.
Zürich
» Details
|
Research Institution
Institut für Informatik Universität Zürich
Zürich
» Details
|
University
University of Zurich
» Details
|
Primary Discipline(s)
Information Sciences
|
Secondary Discipline(s)
Organic Chemistry
|
|
|
High-Performance Parallel Rendering, Computational Science, Scientific Visualization, Large-area Display Wall, parallel rendering
|
|
The continuing improvements in hardware integration lead to ever faster CPUs and GPUs and higher resolution sensor devices, which has enabled major advances in computational sciences and bio-medical imaging, and lead to equally growing complex data sets from simulations and imaging systems. Effective and timely analysis of such vast amounts of data has become a major undertaking, with interactive visualization being one of the most important tools for gaining insight into the structure of data.
However, standard visualization systems cannot efficiently display such exceedingly large data sets, and thus hardware accelerated scalable parallel rendering must be exploited. The benefit of cluster-parallel rendering is that workstation graphics hardware is developing faster than graphics in super-computers, thereby outperforming that integration. A cluster provides scalability to sustain interactive response, supports an arbitrary number of display channels, in contrast to shared memory systems, and is often the source of the data to be displayed, e.g. from numerical simulations. Application domains involving large data visualization include, among others, bio-medical or oil & gas geological imaging, computational chemistry, astrophysics, and computational fluid dynamics, as well as centers for large data analysis and remote visualization, or immersive visualization systems driving display walls, CAVEs and auto-stereoscopic displays for simulation and VR applications.
Previous parallel rendering approaches typically failed in one or more of the following, a) providing only a special domain solution, b) providing a solution that acts as a transparent, but not scalable, abstraction of the graphics layer, or c) providing a solution that requires replacement of the entire existing graphics infrastructure. To date, generic and scalable parallel rendering frameworks that can be adapted to a wide range of experimental and computational science applications in chemistry, physics or mathematics, have yet to be developed. Furthermore, flexible and automatic configurability to arbitrary cluster and display-wall configurations, has not been addressed, but is of immense practical importance to scientists using high-performance interactive visualization as a scientific tool. The above mentioned shortcomings of prior parallel rendering systems are the main focus areas of this project. The technical and algorithmic infrastructure for scalable parallel rendering on cluster systems addressed in this proposal is a major milestone in scientific and high-performance parallel visualization research. |
|
|
|