The Visionarium Lab focuses on the adoption of supercomputing and visual analysis tools to advance science, engineering, and education. The Lab also provides educational and support services to improve access to and adoption of cutting-edge tools that integrate with researchers’ data, questions, and workflows. Like an aquarium or a planetarium, the Visionarium is an instrument to see things you can’t see at home.
First, we installed a 10-Gbps connection to the Virginia Tech Research Network and established baseline of performance by testing the network. This was done with a scripted suite of transfers (of different numbers and sizes) over VT-Rnet between the Visionarium equipment on campus and the ARC supercomputers at the data center. Satisfied that the throughput and latency were within the expected range, we considered the installation phase complete.
One of the flagship systems in the Visionarium Lab is a high-resolution immersive CAVE called the HyperCube. The HyperCube is an 8-projector system that displays 27.6 million stereo pixels at interactive frame rates. Transfers from online sources or ARC storage resources to the HyperCube are greatly improved over the prior 1-Gbps network service. This has been a great benefit to speed up data movement for local rendering, especially for our HyperCube users in Biochemistry, Medical Imaging, and Mechanical Engineering.
Recently, the emphasis has been on the development of big data visualization applications that can take advantage of this connectivity. Visionarium faculty worked with ARC Graduate Research Assistants to develop a high-performance remote visualization infrastructure over VT-Rnet. These GRA students also used these systems in their graduate research in the Department of Computer Science.
A highlight accomplishment is the demonstration and evaluation of Immersive Paraview and using it for remote rendering from ARC HPC to the HyperCube over VT-Rnet. This work was published in the peer-reviewed HPC Symposium and presented in Baltimore, MD in 2018:
Faiz Abidi, Nicholas Polys, Srijith Rajamohan, Lance Arsenault, and Ayat Mohammed. 2018. Remote high performance visualization of big data for immersive science. In Proceedings of the High Performance Computing Symposium (HPC '18). Society for Computer Simulation International, San Diego, CA, USA, Article 5, 12 pages. ISBN: 978-1-5108-6016-2 (ACM DL)
This capability is directly benefiting several faculty and students, actively supporting science areas such as : Astronomy (large galaxy catalogs), Mathematics (dynamic systems), Geosciences (carbon sequestration simulations) and Mechanical Engineering (point clouds, structures, airflows).
Theses and Dissertations
Both of the following students successfully defended their work and graduated in Spring 2017.
- Faiz Abidi (Masters) developed and benchmarked several strategies to feed the HyperCube pixels directly from live sessions on ARC HPC machines. Using open tools like TurbVNC, he was able to successfully connect Paraview running on the cluster to our local projectors and tracking systems at interactive rates. Thesis: “Remote High Performance Visualization of Big Data for Immersive Science”. Faiz joined NetApp.
- Ayat Mohammed (PhD) worked with Paraview and the HyperCube in several of the case studies in her dissertation:
“High-dimensional Data in Scientific Visualization: Representation, Fusion and Difference”. Dr. Mohammed is now a PostDoc at Texas Advanced Computing Center.