Web-based Techniques for Remote Scientific Visualization

The objective of this Phase II proposal is to design and implement a web-based visualization system to aid in the analysis of High Performance Computing (HPC) simulations. HPC simulations are widely used in the design of new materials and weaponssystems. However, the amount of data they generate is large, limiting the analysis of the HPC results to a few high-powered computers. In this proposal we introduce the notion of a visualization server, much like a traditional web server, that wouldsupport the analysis and exploration of HPC simulations from a wide range of clients. The visualization server and clients would be built upon the Visualization ToolKit (VTK), a widely used, comprehensive, portable, open-source visualization library. Thesystem uses secure methods of data transmission requiring only basic functionality from a standard Web browser. Because the system takes advantage of VTK's parallel streaming pipeline, large amounts of data (a petabyte has been documented) can be processedand rendered in parallel. The system will be developed by Kitware, Inc., a small business providing commercial support for VTK, and whose employees are the original creators and maintainers of VTK.The proposed project will significantly improve the use andaccessibility of HPC clusters in many military and commercial fields. HPC is already used for ballistics, ablation, impact analysis, weather prediction, new weapon systems, communications coverage, tracking systems, logistics, nuclear testing,environmental analysis, and computational fluid dynamics. The web-based client application will provide an easy to use interface to a powerful visualization server. This allows analysts with the proper authentication to quickly and easily access HPCresults from their desktop or palmtop computer. The visualization server will make using HPC resources less labor intensive and more powerful by providing a flexible scalable tool that can process data sets of virtually any size on any clusterarchitecture. The proposed tools will be invaluable post-processing and run-time resources for analyzing and debugging large simulations. Specific benefits will include faster analysis cycles with less wasted runs, and easier generation of animationsfor communicating simulation results.