ACADEMIA

MOUNTAIN VIEW, Calif., SGI today announced that the National Center for Atmospheric Research (NCAR) in Boulder, Colo., has installed SGI(R) Visual Area Networking (VAN) technology aimed at ushering in the next generation of very-large-data exploration and collaborative computing. NCAR's new visualization lab is powered by an SGI(R) Onyx(R) 3800 graphics supercomputer running SGI(R) OpenGL Vizserver(TM) software, enabling individual users or teams to access the advanced visualization, high-performance computing and complex data management capabilities of a centralized visualization server. The VisLab, managed by NCAR's Scientific Computing Division (SCD), is a synthesis of people, computers, tools and techniques aimed at a simple goal: to advance the atmospheric and related sciences by direct application of state-of-the-art collaborative visualization. The new facility combines powerful visualization computing platforms, high-bandwidth networking and a high-resolution Barco(TM) stereo 3D-display system to provide an unprecedented environment for remote collaboration and large-scale data exploration. The computational hub of NCAR's VisLab is a complex of SGI systems with VAN technology. Following an extensive evaluation period, SCD acquired an SGI Onyx 3800 high-performance graphics system for a remote visualization pilot project. The software for the project is SGI OpenGL Vizserver, which enables a single SGI(R) Onyx(R) 3000 series graphics supercomputer to distribute visualization sessions to virtually any kind of client. The rendering, compute and I/O power of the SGI Onyx 3000 series systems is made available on any end-user client device, including laptops, PCs, workstations, wireless tablet devices and, in the future, PDAs. "SGI refers to this as Visual Area Networking, and it's an important strategic step toward expanding the virtual walls of our facility. With this technology, NCAR expects to make significant inroads toward deploying a collaborative remote visualization environment, which has the potential to allow NCAR and university researchers to interact simultaneously in real time with groups scattered around the world," said Don Middleton, leader of SCD's Visualization and Enabling Technologies Section (VETS) at NCAR. "With access to 5TB of RAID 5 storage, this new system will leverage our investment in storage area networking technology nicely, allowing very large data sets to be centralized while collaborators, who may be geographically dispersed, can engage in interactive visualization over a local area network or a wide area network." NCAR's VisLab is also an active node in the new electronic collaboration environment called the AccessGrid, a development that has grown out of U.S. Department of Energy efforts to foster geographically distributed scientific research. The AccessGrid is an ensemble of network, computing and other resources that supports group-to-group human interaction. These interactions facilitated by the AccessGrid can take the form of large-scale distributed meetings, collaborative research, seminars, symposia, lectures, tutorials and training. Middleton adds, "The AccessGrid has fundamentally changed the way that we advance our research program and work with each other. Coupling advanced visualization with these collaborative environments is a very exciting opportunity." VETS, now in its third year of existence, is a dynamic and growing program within SCD. It is focused on knowledge development, the primarily post-computational phase where data moves to information, discovery, insight, and communication. The program includes research, development, deployment and state-of-the-art facilities and spans the realms of large-scale data analysis and visualization, Web environments and infrastructure, collaborative environments and collaboratories, grid technology, scientific data portals, and education and outreach. "We are entering into a revolutionary period," Middleton said. "Advances in computational power, storage, high-bandwidth networking, display technologies and the grid constitute the critical mass for a new generation of collaborative visual computing. Our VisLab is aimed at helping to usher in this new era of visual interaction, providing the ability to freely explore vast data spaces and producing media that can communicate the results of these efforts to the scientific community and the general public." Another goal of SCD is to enable the best atmospheric research in the world by providing and advancing high-performance computing technologies. The multi-vendor production supercomputer environment managed by SCD for NCAR includes four SGI supercomputers, including a new 128-processor SGI(R) Origin(R) 3800 system, named "Chinook," installed at NCAR in July 2002. The 128-processor SGI Origin 3800 is one of the computers used to power NCAR's Climate Simulation Laboratory, which supports large-scale, long-running simulations of the earth's climate system including model components that need to be completed in a short calendar period. The new SGI supercomputer is binary compatible with an SGI(R) Origin(R) 2000 series server it replaced and has the same number of processors but twice the clock speed, four times the memory and more than double the disk capacity. Chinook has 128 processors, 64GB of memory, and 580GB of disk space. It is fitted with a HIPPI connection to the NCAR Mass Storage System and Gigabit Ethernet to NCAR networks. The VAN server and the Chinook system are connected to a Storage Area Network (SAN) and CXFS(TM) (clustered XFS(TM)), the industry's fastest shared filesystem. CXFS provides no-compromise data sharing, enhanced workflow and reduced costs in data-intensive environments by eliminating file duplication and the time it takes to move large files over the network. The SAN allows data to be read directly from terabytes of disk instead of being copied over slow networks from the supercomputer to the file server and back to the VAN server. Al Kellie, the director of NCAR's Scientific Computing Division, concludes: "SGI has long been the workhorse of NCAR's visualization efforts and we look forward to continued progress in getting visualization resources into the hands of the scientists."