HR: 1330h
AN: ED32C-1212 [PDF]
TI: Grid-based Visualization Framework
AU: * Thiebaux, M
EM: thiebaux@isi.edu
AF: USC Information Sciences Institute, 4676 Admiralty Way,
Suite 1001, Marina del Rey, CA 90292 United States
AU: Tangmunarunkit, H
EM: hongsuda@isi.edu
AF: USC Information Sciences Institute, 4676 Admiralty Way,
Suite 1001, Marina del Rey, CA 90292 United States
AU: Kesselman, C
EM: carl@isi.edu
AF: USC Information Sciences Institute, 4676 Admiralty Way,
Suite 1001, Marina del Rey, CA 90292 United States
AB:
Advances in science and engineering have put high demands on tools for high-performance large-scale visual data exploration
and analysis. For example, earthquake scientists can now study earthquake phenomena from first principle physics-based
simulations. These simulations can generate large amounts of data, possibly high spatial resolution, and long time series.
Single-system visualization software running on commodity machines cannot scale up to the large amounts of data generated by
these simulations. To address this problem, we propose a flexible and extensible Grid-based visualization framework for
time-critical, interactively controlled visual browsing of spatially and temporally large datasets in a Grid environment. Our
framework leverages Grid resources for scalable computation and data storage to maintain performance and interactivity with
large visualization jobs.
Our framework utilizes Globus Toolkit 2.4 components for security (i.e., GSI), resource allocation and management (i.e.,
DUROC, GRAM) and communication (i.e., Globus-IO) to couple commodity desktops with remote, scalable storage and computational
resources in a Grid for interactive data exploration. There are two major components in this framework---Grid Data Transport
(GDT) and the Grid Visualization Utility (GVU). GDT provides libraries for performing parallel data filtering and parallel
data exchange among Grid resources. GDT allows arbitrary data filtering to be integrated into the system. It also facilitates
multi-tiered pipeline topology construction of compute resources and displays. In addition to scientific visualization
applications, GDT can be used to support other applications that require parallel processing and parallel transfer of partial
ordered independent files, such as file-set transfer.
On top of GDT, we have developed the Grid Visualization Utility (GVU), which is designed to assist visualization dataset
management, including file formatting, data transport and automatic byte swapping. GVU also supports parameterized data
reduction filters such as point sampling with scalar range culling, as well as volume cropping, and down sampling. The GVU
framework can be used to facilitate the parallel execution of existing transformation filters, such as the VTK marching cubes
iso-surface filter, as well as other custom domain-specific filters. Our initial implementation supports remote synthesis
of view point independent display lists. This feature allows the local display machine to control the view point for reduced
view point latency, and multi-view rendering (e.g., stereo rendering).
The poster will provide an architectural design of the Grid-based visualization framework and its components. We will also
show a brief demo/movie of our work, and how it is being used to visualize 4D earthquake wave propagation data.
UR: http://www.globus.org
DE: 6605 Education
SC: Education and Human Resources [ED]
MN: 2003 Fall Meeting