| Interactive Remote Visualization across the Louisiana Optical Network (LONI) and the National LambdaRail (NLR) | ||
jump to [ description | participants | infrastructure | documentation ] page partially updated january 2006 |
| description |
|
We have demonstrated how high-speed networks can be coupled to compute
resources, data storage archives, and visualization services
to enhance the study
of complex scientific problems such as
the modeling of black holes and other sources of gravitational waves.
Easy to use and flexible capabilities are required for the exploratory
visualization of
large data sets, where scientists need to browse and analyze new data,
looking for new and insightful features.
The most important requirement for such explorations is that any
visualization system
is interactive and responsive -- which may lead to different or even
conflicting design
issues to systems focusing on high throughput.
The scenario (see Figure 1), used a visualization server located at LSU, Baton Rouge, to process data located at remote sites on the new Louisiana state network (LONI) and Europe. The data source was a precomputed, distributed, 120 Gbyte data set generated by a black hole simulation performed using the Cactus Framework (http://www.cactuscode.org). We visualized the 3D data, using direct 3D volume rendering. while at the showfloor, in Baton Rouge and in Brno tangible devices were be used to interact with the visualization server. This demonstration addressed numerous key issues for high speed optical networks: dynamic application driven reservation of lambdas, human-computer devices for effective and intuitive interactions of scientists with their data, mechanisms for efficiently moving data on and off a high capacity pipe, and the development of interactive and collaborative visualization scenarios which can be productively used by scientists. |
| participants | ||||||||||||||||||
|
| infrastructure |
| hardware, software |
|
Data Servers: The data servers, located at selected LONI sites
held the data to be visualized in main memory, requiring in
excess of 100GB total RAM. The data servers were connected to the visualization
server via a high-speed network.
Visualization Server: High end server, located at CCT at LSU, Baton Rouge, with 8 Gbyte RAM, 512 Mb video memory, high speed processor. Visualization was performed using Amira (http://amira.zib.de). Interactive Devices: Tangible devices for interaction were deployed at the demonstration site. (http://www.zib.de/ullmer/jb2002/). |
| network |
|
This demonstration used the following networks:
LONI: The Louisiana Optical Network Initiative (LONI) will provide a multiple lambda 40Gbps link between CCT/LSU, Louisiana Tech, University of New Orleans, Southern University and University of Louisiana at Lafayette. LONI will also include multiprocessor compute resources at each site (except for LSU which has existing resources). NLR: The National LambdaRail provided a lambda link between LONI (node at Baton Rouge) and Chicago. International Links: To reach Brno from Chicago, we used the Starlight, NetherLight and CzechLight connections.
|
| documentation |
| articles, presentations |
|
iGrid 2005 Presentation
Supercomputing 2005 Presentation "Distributed and collaborative visualization of large data sets using high-speed networks" [preprint]: Andrei Hutanu, Gabrielle Allen, Stephen D. Beck, Petr Holub, Hartmut Kaiser, Archit Kulshrestha, Milo Liska, Jon MacLaren, Ludek Matyska, Ravi Paruchuri, Steffen Prohaska, Ed Seidel, Brygg Ullmer, Shalini Venkataraman |
| software |
|
Grid Application Toolkit (version 1.6.2 used for the demonstration) |
