This project addresses what is fast becoming the most common, most severe bottleneck in high-end computational science: the inability of users to analyze their data in real time because of the limitations of their local systems and limited network bandwidth that cannot support transfer of tremendous data sets. The project will bring data analysis and visualization capability in line with HPC modeling and simulation capability to enable new kinds of interaction, increasing the likelihood of discovery.
The Challenge: Scientific visualization is a fundamental data analysis technique for simulation-based research. Through 2-D and 3-D images, scientific visualization helps scientists explore, make sense of, and communicate data, whether it is modeling a hurricane, tracing the arterial blood flow of a heart, or exploring a super massive black hole. However, high performance computing (HPC) systems capabilities are racing far ahead of users' ability to effectively visualize the data they produce.
As petaflop systems produce simulation output of unprecedented scale, it is becoming unfeasible to move these enormous data sets over networks to visualization systems and to use special-purpose visualization systems to interact with them. To fully achieve the scientific impact of these costly tera-and petascale systems, we need to improve the visualization capabilities for high end users. As with HPC, this requires scalable visualization tools that aggregate the capabilities of many compute nodes, while rendering the data close to the source to eliminate costly network transfers. This requires remote visualization interfaces to these large-scale visualization systems, enabling remote users to work interactively with their data.
The Solution: The development of the Scalable Collaborative and Remote Visualization Software (SCoReViS) is a direct response to the need for next-generation visualization tools for large-scale HPC platforms and dedicated graphics clusters. SCoReViS will: o Leverage the computational power of tera- and petascale HPC and large GPU-based graphics clusters, and scale seamlessly from the largest available platforms down to smaller systems deployed at scientists' local institutions; o Maximize the investments in large-scale systems by providing remote and collaborative access to visualization tools, making them available and effective for any researcher with a reasonably high-bandwidth network connection (e.g. consumer broadband service); o Provide general purpose visualization capabilities by supporting any software based on the standard OpenGL API, thus enabling most popular visualization applications; and o Create a platform of tools for scientists who compute at the largest scale and to promote the development of tools to address issues associated with large, time-dependent data sets. The key components will be assembled, optimized, packaged and supported to provide high performance, scalable remote/collaborative visualization on petascale HPC and graphics clusters.
Impact: The potential impact of this project is immeasurable. The resulting software will be made available via open source licensing, enabling use by anyone with access to the TeraGrid, including diverse scientific disciplines. The remote and collaborative visualization capabilities enabled by SCoReViS will allow access by a broader community of researchers, increasing the ability of high performance computing (HPC) to address the largest problems facing us.
