|
Titel |
Visualization in a Climate Computing Centre |
VerfasserIn |
Karin Meier-Fleischer, Niklas Röber, Michael Böttinger |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250098241
|
Publikation (Nr.) |
EGU/EGU2014-13902.pdf |
|
|
|
Zusammenfassung |
Today, the extensive numerical simulations of climate models require elaborate visualization
for understanding and communicating the results. Typical data sets of climate models are
3-dimensional, multivariate and time dependent, and can hence be very large. Interactive
visual data analysis improves and accelerates the comprehension of these vast amounts of
data.
At DKRZ, the German Climate Computing Centre, a central high end visualization
server, various domain specific visualization applications, and a remote 3D rendering solution
enable users to interactively visualize their extensive model results right at their
desktops.
The DKRZ’s visualization server is a heterogeneous Linux cluster, currently consisting of
10 state of the art visualization nodes equipped with 96 -256 GB RAM and high end NVidia
GPUs. Since the parallel file system of the DKRZ’s supercomputer is directly mounted over a
powerful network, the model data can directly be analyzed and visualized. VirtualGL and
TurboVNC are used for utilizing the server’s GPUs for 3D rendering, while the TurboVNC
client on the user’s local computer continuously displays the resulting video stream. By using
this central visualization server instead of a local computer, three main benefits are
achieved:
Time consuming transfers of large data sets from the supercomputer to the local
computer are not needed.
The hardware of the user’s local workstation doesn’t need to be powerful, no
expensive GPU is required.
Users don’t have to install or buy visualization software.
On the visualization server, a wide range of visualization software is installed. Avizo Green, a
powerful commercial software customized for interactive 3D visualization of climate model
data, is available, as well as SimVis and ParaView, which focus more on an exploratory
visualization of data. SimVis and ParaView provide techniques like Linking & Brushing to
emphasize or de-emphasize portions of the data. Furthermore, some domain specific 2D
graphics software packages, like NCL and GrADS, as well as software for processing,
manipulating and analyzing the data, such as the CDOs (Climate Data Operators), are also
used on the DKRZ visualization server.
This PICO will give an overview on the overall system and the techniques applied at
DKRZ for the visualization of climate modeling results. Many examples are given to
illustrate the types of applications. |
|
|
|
|
|