VISUALIZATION

By Aaron Dubrow & Faith Singer-Villalobos -- TACC’s multi-user environment allows users to see science on an unprecedented scale: Whether researchers are modeling microscopic images of the smallest molecular structures, or astronomical simulations of the formation of the universe, numbers alone tell only part of the tale. Scientific visualization is therefore an essential technique for transforming the raw data from simulations and from observations into information and knowledge about the world around us. However, as supercomputers such as Ranger grow in power, they can perform simulations of such amazing size and detail that they cannot be represented visually by a typical computer and display without losing resolution and scale. Consequently, it has become crucial to add another component to TACC’s advanced computing capabilities, so that researchers can more easily interact with the fine-grained detail embedded in massive datasets — an environment for visualizing data and ‘seeing’ science on an unprecedented scale. “The process of scientific visualization transforms raw data into a logical, visual format allowing scientists in practically every field of science, from aerospace engineering to astrophysics to archeology, to see hidden aspects of their research and make assertions they hadn’t considered before,” said Kelly Gaither, Associate Director for Data and Information Analysis at the Texas Advanced Computing Center (TACC). The lab's "crown jewel," Stallion, is a 20.5’ x 7.5’ display environment comprising 45 Dell 30-inch flat panel monitors offering 184 million pixels, powered by a cluster of 23 high-end Dell XPS systems and a Dell Precision 690 head node. On October 10, 2008, TACC, one of the leading advanced computing centers in the country, unveiled the ideal technology environment for the job: a modern, reconfigured Visualization Laboratory (VisLab) capable of reproducing terascale (one trillion bytes) data sets with exceptional clarity and resolution. “People absolutely love clear, crisp pixels and larger scale displays at a higher resolution,” Gaither said. “Those are two things that the user community can take advantage of today. You have to have a reason for people to get up out of their lab and come to a new environment. We’ve really tried to address what the science community needed.” The 2,900-square-foot lab, located on the ground floor of the Advanced Computational Engineering and Sciences building at The University of Texas at Austin, houses Stallion, a 20.5’ x 7.5’ display environment comprising 45 Dell 30-inch flat panel monitors offering 184 million pixels, powered by a cluster of 23 high-end Dell XPS systems and a Dell Precision 690 head node. This configuration allows an extremely high level of detail and quality compared to a typical moderate pixel count projector. With 47 NVIDIA graphics cards, 108 gigabytes of system memory, and more than 36 gigabytes of graphics memory, Stallion enables datasets to process on a massive scale, and allows the interactive visualization of substantial geometries. Paul Navratil, one of TACC’s visualization scientists, oversaw the lab’s construction and helped integrate the new technology. “It’s quite literally about getting deeper into the data,” he said. “The data sets produced by Ranger are at a fidelity that has never been produced before, but you lose some of the power of that fidelity if you’re only able to display it on a single 30-inch display. You can only show something down to the resolution of a single pixel, so the more pixels you have, the more data you can display across the image.” By means of comparison, an average desktop displays roughly 1600x1200 pixels of resolution, or under two megapixels — Stallion has nearly 100 times times more resolution. In addition to Stallion, the lab also now offers a state-of-the-art Sony ultra high resolution projector, capable of displaying crisp imagery at a massive size of 20 feet across by 10 feet high; four high-performance workstations for simulating and analyzing data to eventually view on the larger displays; and a glass-walled conference room with remote meeting and presentation capabilities. All of these visualization resources are linked through a shared file system, so researchers can easily move data from one resource to another. The new incarnation of the VisLab marks a dramatic departure from previous efforts, and reflects a changing philosophy about constructing an environment for displaying science on a large scale. Kelly Gaither, Associate Director for Data and Information Analysis at the Texas Advanced Computing Center (TACC) “The previous instantiation of the lab was based on high-end, integrated technology,” explained Gaither, who is also a co-designer of the space. “It was very expensive and made visualization an elite activity. Now, the lab is built with commodity technology, making it available to anybody.” The former VisLab technologies cost approximately $5 million to assemble in 2001. Thanks in part to a collaborative grant from Dell, Intel, Microsoft and Cisco, TACC deployed the new technologies for roughly $500,000, providing 10 times the performance of the previous space at just one-tenth of the cost. And since the lab uses relatively inexpensive commodity parts, it can be easily upgraded with new processors, commodity graphics cards, and projectors to avoid obsolescence. Another major change made by the TACC design team impacted the configuration and usage of the lab. The previous space was designed as a single-user environment, and only one project could be worked on at a time. The new lab has four visualization areas that can be used together or separately, greatly enhancing the value and efficacy of the lab and encouraging collaboration. Gaither sees this lean, flexible arrangement as a potential model for other universities and computing centers that have come to believe that vislabs might be too expensive of an investment. Though it’s too soon to tell what discoveries will arise from these new visualization capabilities, developers at TACC, who have seen advancing technology continually drive innovative science, expect great things. “Working with [UT astrophysicist] Volker Bromm on the dissolution of early stars, our visualization revealed that the process is heterogeneous, and stars deteriorate in a non-uniform way,” Navratil said. “That was a very interesting result for them, showing that often the small-scale structures can be quite meaningful.” Though experienced researchers in hard sciences like cosmology, physics, geology and engineering — the typical users of high-performance computing — will be among the main users of the new resources, the lab is designed for a much broader audience. Says Gaither, “I think it’s definitely going to be a valuable outreach tool whether it’s for K-12 or it’s a primary draw for the ‘Blue Ribbon’ tours at UT. The K-12 community is going to love the Sony screen because it’s a seamless display...it feels more like a movie theater. It bridges the gap and makes science a little ‘cooler.’” Navratil agrees, “Once a tool is provided to someone, they’ll find new and innovative ways to use it. So having a facility of this sort, I’m thrilled to see what non-traditional researchers might do with it.” The Sony ultra high resolution projector, capable of displaying crisp imagery at a massive size of 20 feet across by 10 feet high, shows an image of a WRF weather forecast. To those who could get lost in the technology, Gaither reiterated that the new lab is more than just pixels and processors. One of its greatest strengths, she said, is the expertise and support available to users. “Manning the lab, making sure that when people come in they have someone to talk to so they’re not just walking into the technology — that sets us apart,” Gaither said. “This is definitely new and innovative. It’s a catalyst to go to the next level of science,” Gaither said. “Much in the same way that TACC’s supercomputers have been a catalyst for going to a new level for scientific simulations, the lab will be a catalyst for developing new kinds of visualization as well. The TACC Visualization Laboratory is available on a reservation basis to researchers at The University of Texas at Austin, the UT System, as well as to members of the TeraGrid community. For information on tours and training, please contact Makoto Sadahiro, Visualization Laboratory Manager at: makoto@tacc.utexas.edu or 512.475.6954.