APPLICATIONS

HP today announced a development that allows high-performance computing customers to interactively render and visualize the huge data sets created by complex simulations and models. By adding the HP Parallel Compositing Library to its Scalable Visualization Array offering, the company has tapped the inherent performance scalability of clusters for network-based pixel compositing. The combined technology will greatly enhance the visualization capabilities for complex scientific and engineering applications such as seismic analysis, engineering design and simulation, medical imaging, weather models and animation that demand more performance than a single graphics system can deliver. "The amount and complexity of our scientific data is expanding exponentially and the interactive rendering of these huge data sets requires exploiting parallelism," said Dr. Jean M. Favre, chief scientist and leader of the scientific visualization group at CSCS, the Swiss National Supercomputing Centre. "With the use of the HP SVA graphics clusters and the HP Parallel Compositing Library, these data sets can now be rendered efficiently and effectively, allowing us to better visualize and gain insight into our scientific simulations and models, such as proteins and molecular structures." The HP Parallel Compositing library is a tuned, documented and supported implementation of the Parallel Compositing application programming interface, which was developed by HP over the last several years with input from industry thought leaders such as Computational Engineering International (CEI), CSCS, BME (Budapest University of Technology and Economics), and the University of Manchester, U.K. The library includes a number of enhancements, such as better error handling, a network layer tuned for InfiniBand and a user's guide, and it is designed for flexibility in that it fits with a wide variety of application structures while not imposing any structure on the applications. The library consists of a set of related functions that make it possible for multiple graphics nodes in a cluster to collectively produce images. By distributing the load over multiple systems, the library enables the processing of significantly more graphics and larger data sets and can create larger images than any individual graphic card. To accomplish this, the library produces a series of images called "frames" in real time, which involves graphic cards producing "sub-images" that make up parts of the final image. Using a powerful scaling technique called "sort-last parallel rendering," compositing operators then combine these sub-images to produce the final image. "The absence of a common enabling interface has been an impediment to exploiting the power of parallel graphics in cluster environments," said Winston Prather, vice president and general manager, High Performance Technical Computing, HP. "HP teamed with industry leaders like CEI to advance the technology, and we are committed to delivering optimum results for customers through continuous innovation." The HP Parallel Compositing Library is available with the HP Scalable Visualization Array, a highly affordable, scalable, ready-to-run visualization solution that completes the HP Unified Cluster Portfolio's integration of computation, data management and visualization in a single, integrated cluster environment. The open Parallel Compositing API specification is available on the HP Collaboration and Competency Network website at www.hp.com/techservers/hpccn/downloads/pc_specification.pdf. More information on the HP Scalable Visualization Array and high-performance computing from HP is available at www.hp.com/go/visualization and www.hp.com/go/hptc, respectively.