The exhibit will demonstrate hardware and software solutions for visualizing extremely large datasets. ParaView, a parallel visualization tool developed by Kitware Inc. and Los Alamos as part of the ASCI VIEWS program, will visualize the results of real-world scientific simulations on a commodity visualization cluster running Windows 2000. The five node visualization cluster demonstrated at Supercomputing is representative of a 128-node visualization cluster currently being prototyped at Los Alamos National Laboratory.
The Los Alamos Computer ARchitecture Toolkit (a la carte) project, also being demonstrated, uses scientific visualization techniques to help analyze the implications of scaling parallel supercomputing architectures.
The a la carte project has simulated 64 node and 4096 node architectures connected by switches arranged in a fat tree. The visualization tools under development help the team to understand the architecture of this network of processors and the dynamics of the virtual circuits being established and the messages being passed between them during the transmission of simulated loads running on the simulated machine.
These visualization tools help the team debug their simulation as well as to recognize and understand possible communication bottlenecks, resource mismatches and other anomolous features of these systems. These same tools should be applicable to real, running environments with similar architectures.
These tools were created in conjunction with the AHPCC at the University of New Mexico using the Flatland visualization development environment built by them.