Publications

Results 26–31 of 31

Search results

Jump to search filters

Integrating software architectures for distributed simulations and simulation analysis communities

Linebarger, John M.; Fellig, Daniel F.; Moore, Patrick C.; Hawley, Marilyn F.; Sa, Timothy J.

The one-year Software Architecture LDRD (No.79819) was a cross-site effort between Sandia California and Sandia New Mexico. The purpose of this research was to further develop and demonstrate integrating software architecture frameworks for distributed simulation and distributed collaboration in the homeland security domain. The integrated frameworks were initially developed through the Weapons of Mass Destruction Decision Analysis Center (WMD-DAC), sited at SNL/CA, and the National Infrastructure Simulation & Analysis Center (NISAC), sited at SNL/NM. The primary deliverable was a demonstration of both a federation of distributed simulations and a federation of distributed collaborative simulation analysis communities in the context of the same integrated scenario, which was the release of smallpox in San Diego, California. To our knowledge this was the first time such a combination of federations under a single scenario has ever been demonstrated. A secondary deliverable was the creation of the standalone GroupMeld{trademark} collaboration client, which uses the GroupMeld{trademark} synchronous collaboration framework. In addition, a small pilot experiment that used both integrating frameworks allowed a greater range of crisis management options to be performed and evaluated than would have been possible without the use of the frameworks.

More Details

Evolving the Web-Based Distributed SI/PDO Architecture for High-Performance Visualization

Holmes, Victor P.; Linebarger, John M.; Miller, D.J.; Vandewart, Ruthe L.

The Simulation Intranet/Product Database Operator (SI/PDO) project has developed a Web-based distributed object architecture for high performance scientific simulation. A Web-based Java interface guides designers through the design and analysis cycle via solid and analytical modeling, meshing, finite element simulation, and various forms of visualization. The SI/PDO architecture has evolved in steps towards satisfying Sandia's long-term goal of providing an end-to-end set of services for high fidelity full physics simulations in a high-performance, distributed, and distance computing environment. This paper describes the continuing evolution of the architecture to provide high-performance visualization services. Extensions to the SI/PDO architecture allow web access to visualization tools that run on MP systems. This architecture makes these tools more easily accessible by providing web-based interfaces and by shielding the user from the details of these computing environments. The design is a multi-tier architecture, where the Java-based GUI tier runs on a web browser and provides image display and control functions. The computation tier runs on MP machines. The middle tiers provide custom communication with MP machines, remote file selection, remote launching of services, load balancing, and machine selection. The architecture allows middleware of various types (CORBA, COM, RMI, sockets, etc.) to connect the tiers depending upon the situation. Testing of constantly developing visualization tools can be done in an environment where there are only two tiers which both run on desktop machines. This allows fast testing turnaround and does not use compute cycles on high-performance machines. Once the code and interfaces are tested, they are moved to high-performance machines, and new tiers are added to handle the problems of using these machines. Uniform interfaces are used throughout the tiers to allow this flexibility. Experiments test the appropriate level of interface: either a large set of specific function calls or a small set of generic function calls. This architecture is based on the goals and constraints of the environment: huge data volumes (that cannot be easily moved), use of multiple middleware protocols, MP platform portability, rapid development of the visualization tools, distributed resource management (of MP resources), and the use of existing visualization tools.

More Details

Distributed design tools: Mapping targeted design tools onto a Web-based distributed architecture for high-performance computing

Holmes, Victor P.; Linebarger, John M.; Miller, D.J.; Poore, Clark A.; Vandewart, Ruthe L.

Design Tools use a Web-based Java interface to guide a product designer through the design-to-analysis cycle for a specific, well-constrained design problem. When these Design Tools are mapped onto a Web-based distributed architecture for high-performance computing, the result is a family of Distributed Design Tools (DDTs). The software components that enable this mapping consist of a Task Sequencer, a generic Script Execution Service, and the storage of both data and metadata in an active, object-oriented database called the Product Database Operator (PDO). The benefits of DDTs include improved security, reliability, scalability (in both problem size and computing hardware), robustness, and reusability. In addition, access to the PDO unlocks its wide range of services for distributed components, such as lookup and launch capability, persistent shared memory for communication between cooperating services, state management, event notification, and archival of design-to-analysis session data.

More Details

Dynamic isosurface extraction and level-of-detail in voxel space

Linebarger, John M.

A new visualization technique is reported, which dramatically improves interactivity for scientific visualizations by working directly with voxel data and by employing efficient algorithms and data structures. This discussion covers the research software, the file structures, examples of data creation, data search, and triangle rendering codes that allow geometric surfaces to be extracted from volumetric data. Uniquely, these methods enable greater interactivity by allowing an analyst to dynamically specify both the desired isosurface threshold and required level-of-detail to be used while rendering the image. The key idea behind this visualization paradigm is that various levels-of-detail are represented as differently sized hexahedral virtual voxels, which are stored in a three-dimensional kd-tree; thus the level-of-detail representation is done in voxel space instead of the traditional approach which relies on surface or geometry space decimations. This algorithm has been implemented as an integral component in the EIGEN/VR project at Sandia National Laboratories, which provides a rich environment for scientists to interactively explore and visualize the results of very large-scale simulations performed on massively parallel supercomputers.

More Details
Results 26–31 of 31
Results 26–31 of 31