Join us at Super Computing 2011!We invite you to visit us at the SC2011 conference in Seattle, Nov 14-17 at Booth #2040 See first-hand how we are enabling research discovery with Dell HPC solutions.
Introduction by Dell's Dr. Jeff Layton
Everyone is probably aware that the amount of computational power in HPC is growing at a tremendous rate. Problem sizes are growing, fidelity of the models is improving, more difficult problems are being tackled, and even problems that involve multi-disciplines are being attacked with computational tools. But producing so much data has its downside – how to “mine” the data for understanding and knowledge?One way to search for understanding in the data is to visualize it. That is, taking the data and converting it to a form that you can actually examine with your eyes. The human mind is an amazing pattern recognition engine so having a visual representation of the information or the mind to “chew on” is a fantastic way to gain insight into computational results. But as mentioned, problems are getting HUGE! So how do you take a problem with TB’s of data, create a useful visual image, and display it for people to use?Conventional monitors such as those on the desktop have limited resolution so you may miss some important details of the results. So what many research groups, companies, and universities are doing is to create large-scale, high-resolution, visualization systems. What you are about to read is blog from Mr. Paul Navrátil at the University of Texas Advanced Computing Center (TACC) describing their efforts in creating the tools for REALLY large-scale visualization. TACC has the highest resolution visualization wall in the world and is investing in a new set of hardware that couples visualization and data analysis for remote users.As you read Mr. Navrátil’s blog pay attention to what he’s describing. Then ask yourself some simple questions:
You don’t have to replicate the systems that TACC has created, but understanding the how and the why TACC is attacking this important area can be very important. Plus, the systems that Mr. Navrátil describes are scalable so you can start smaller and work up to “TACC-scale” if you need. -- Dr. Jeff Layton
It’s an exciting time for large-scale visualization at TACC! In conjunction with SC08, nearly a year and a half ago, we launched the world’s highest resolution tiled display: Stallion, a 307 Megapixel display and 24 node visualization cluster. During this time, Stallion has displayed a wide assortment of breath-taking scientific imagery, including:
We have also featured a number of high-resolution photographs by Austin photographer Ricardo Meleschi, perhaps the most spectacular of which is a 521 Megapixel image of downtown Austin at night.
Stallion has been featured in countless presentations, including a doctoral defense from the Jackson School of Geosciences that used tiled images, video streams and simulations on Stallion to create an interactive research mural.Then, in January 2010, we put a new system into production: Longhorn, a 2048 core, 512 GPU cluster designated for remote visualization and data analysis (VDA), the largest dedicated VDA machine in the world. By using Dell Precision R610 and R710 nodes, Longhorn contains a powerful combination of computation (8 Intel ‘Nehalem’ cores per node), memory (at least 48 GB per node) and hardware acceleration (2 NVIDIA Quadro FX5800 GPUs per node). This machine opens new possibilities for remote and collaborative visualization. We have visualized enormous datasets, ranging from massive simulations of the Universe to the largest isotropic turbulence simulations. In addition, we have streamed interactive visualizations from Longhorn to displays as far away as Portland, OR (on the SC09 show floor) and Queensland, Australia. Further, Longhorn has become a platform for cutting-edge research in distributed GPU computing, such as the hybrid molecular dynamics codes of Dr. David LeBard that can achieve 10x speed-up on Longhorn over his code running on TACC’s Ranger CPU-only supercomputer (currently #8 in the Top500 list).
There is excellent work and exciting opportunities at TACC, and in this space, we will regularly share some of that work with you and discuss some of the technology that powers the work. We hope you’ll find it useful, motivating and as exciting as we do.What are your thoughts? We'd love to hear your comments on this very important HPC subject.
Paul A. Navrátil biographical Sketch Paul is a Visualization Scientist in the Data and Information Analysis division of the Texas Advanced Computing Center (TACC) at the University of Texas at Austin. His research interests include efficient algorithms for large-scale parallel visualization and data analysis (VDA) and innovative design for large-scale VDA systems. His recent work includes algorithms for ultrascale distributed-memory ray tracing, work that enables photo-realistic rendering of the largest datasets produced on supercomputers today, such as cosmologic simulations of the Universe and computational fluid dynamics (CFD) simulations at unprecedented levels of detail. Paul also helps manage TACC’s visualization systems, including Longhorn, the largest supercomputer in the world dedicated to VDA, and Stallion, the highest-resolution tiled display in the world. Paul received a B.S. in Computer Science, a B.A. in Plan II interdisciplinary honors, and an M.S. in Computer Science from the University of Texas at Austin. He will complete a Ph.D. in Computer Science from UT-Austin in May, 2010. Paul’s research has been published in the IEEE Transactions on Visualization and Computer Graphics and his visualizations have appeared in The New York Times and Discover magazine.