The HP Scalable Visualization Array (HP SVA) launched at the SC|05 Supercomputing Conference leverages industry-standard components and advances in clustering, graphics and networking technology to provide scientists and engineers with the affordable, flexible and powerful visualization and analytical capabilities they need to solve increasingly complex problems.
A key part of the HP Unified Cluster Portfolio, HP SVA is a scalable visualization solution that, when used with other portfolio components, enables the integration of computation, data management and visualization in a single cluster environment, greatly enhancing productivity in HPC environments.
Visualization is a critical capability to enhance the productivity of HPC environments. High-performance scientific and engineering problems generate very large data sets that need to be visualized, often interactively with computational jobs. Until now, available solutions did not meet this need because computational performance outpaced capabilities in visualization tools and available solutions were too expensive and/or limited in performance and scalability.
“Our visualization applications rely on a very robust multi-node platform provided by HP to deal for the first time with many gigabytes of time-dependent data and screen resolution of 3840x3072 pixels for our Image Wall,” said Dr. Jean M. Favre, chief scientist and leader of the scientific visualization group at CSCS, the Swiss national supercomputing centre. “HP SVA is a logical step forward in providing inexpensive, powerful solutions that work for users in many ways and make it possible for more universities and supercomputing centers to acquire the visualization tools they need.”
CSCS currently runs an Intel® Itanium®-based HP cluster used for data serving, coupled with bi-directional node-to-node connections with a cluster of HP workstations doing parallel graphics rendering.
HP SVA leverages state-of-the-art industry-standard and open source clustering, graphics and networking technology to reduce costs and enhance flexibility. The tight integration of scalable computation, data management and visualization enables:
- clustered parallel visualization applications with support for very large data sets;
- display of complex, high resolution images, including volume visualization; and
- real-time rendering, with computational steering via closely coupled visualization, computation and data management.
The HP SVA consists of a cluster of HP workstations in building block configurations, open source Linux software, industry-standard graphics cards and network adaptors, and an integrated software system. System software includes HP XC System Software for cluster infrastructure and workload management, and HP Scalable Visualization Array software for configuration and job management. HP StorageWorks Scalable File Share (HP SFS) software and HP Remote Graphics software are optional.
Visualization and graphics libraries are provided by third-party vendors and the open source community. Commercial applications and applications written by end users can run on SVA, taking full advantage of the various toolkits and libraries. HP SVA is also compatible with all Open Graphics Applications Programming Interfaces (APIs).
Scalability and flexibility
The SVA scales to support diverse visualization workloads including multi-user, multi-tasking and multi-sessions. Cluster users can add as many visualization nodes to their environment as they require. These nodes are shared resources and can be used together for scalable performance or individually by different applications. The resulting visual images can be viewed remotely on any standard desktop system.
HP SVA supports various visualization styles, models and display systems including single screens, caves or walls. The SVA technology can produce a vast, high-resolution display wall of 100 million pixels and more.
HP SVA works in three flexible modes that can be mixed and dynamically changed as requirements change, using the SVA job and resource management tools. The three modes are:
- A cluster of independent workstations with centrally managed resources for easy scheduling and running any application, increasing workstation utilization and improving ROI;
- A cluster of synchronized workstations with the capability to scale across a number of displays with middleware offering spatial compositing for applications;
- A sort-last computing cluster that is a Linux cluster visualization system that distributes image data among workstations working in parallel to render a portion of the image faster, then composited to produce the final image.
SVA partners
In addition to making high-performance visualization affordable and accessible, customers require the availability of applications that can take advantage of the visualization cluster technology. Applications including Wolfram Research’s gridMathematica and CEI’s EnSight offer users the performance and comfort in working with well-known, trusted applications.
“Wolfram is excited to work with HP on scalable visualization, as gridMathematica users have been exploring areas where increased resolution is mandatory,” said Elises Cervantes-Pimental, senior kernel developer of Wolfram Research, Inc., a leading technical software company.
“CEI is in the process of qualifying EnSight DR on the HP Scalable Visualization Array,” said Randall Frank, senior developer, CEI, a leading provider of engineering and scientific graphics software. “The emergence of large, high-fidelity simulation methods on distributed memory HPC systems has necessitated a transition of visualization applications to rendering paradigms that embrace higher resolution image generation and unprecedented increases in software scalability. EnSight DR, its accompanying distributed image compositor, graphics-enabled PC clusters and other distributed system technologies specifically address the needs of CEI customers who are utilizing such simulation methods on an increasingly regular basis.”
HP market leadership
Driven by the availability and flexibility of its product portfolio, HP continues to lead the HPC market with more than 30 percent market share according to second quarter 2005 figures released by IDC.(1)
HP systems also maintain the top two rankings in Purdue University’s Top Application Performers (TAP) list, which ranks high-performance computers based on their performance running the SPEC HPC2002 benchmark suite. The HPC2002 suite consists of large-scale computer applications, including computational chemistry, climate modeling and seismic processing.
About HP
HP is a technology solutions provider to consumers, businesses and institutions globally. The company’s offerings span IT infrastructure, global services, business and home computing, and imaging and printing. For the four fiscal quarters ended July 31, 2005, HP revenue totaled $85.2 billion. More information about HP (NYSE, Nasdaq: HPQ) is available at www.hp.com.
(1) Worldwide Technical Server 2Q05 Vendor Shares, IDC.
Intel and Itanium are registered trademarks of Intel Corp. or its subsidiaries in the United States and other countries. Linux is a U.S. registered trademark of Linus Torvalds.
|
Printable version
