Authors:
Tor-Magne Stien Hagen
;
Daniel Stødle
and
Otto J. Anshus
Affiliation:
University of Tromsø, Norway
Keyword(s):
Visualization, High-resolution tiled display walls, Live data sets, On-demand computation.
Related
Ontology
Subjects/Areas/Topics:
Algorithms and Technologies
;
Computer Vision, Visualization and Computer Graphics
;
Data Acquisition and Information Processing
;
General Data Visualization
Abstract:
Visualization of large data sets on high-resolution display walls is useful and can lead to new discoveries that would not have been noticeable on regular displays. However, exploring such data sets with interactive performance is challenging. This paper presents live data sets, a scalable architecture for visualization of large data sets on display walls. The architecture separates visualization systems from compute systems using a live data set containing data customized for the particular visualization domain. Experiments conducted show that the main bottleneck is the compute resources producing data for the visualization side. When all data is cached in the live data set, the main bottleneck (decoding images to create OpenGL textures and constructing geometry from raster data) is on the visualization side. On a 22 megapixel, 28 node display wall, the visualization system can decode 414.2 megapixels of images (19 frames) per second. However, the decoding is multi-threaded, and inc
reased performance is expected using multi-core computers.
(More)