The Computer Sciences group provides the majority of the research and development related to scientific computing at the California site. Our mission is to underpin Sandia/California’s current business areas with robust and exciting research capabilities and to provide fundamental knowledge for future innovation.
The Interactive Design Center at Sandia/California
The Computer Sciences and Information Systems (CS&IS) Center supports two outstanding research facilities, the Distributed Information Systems Lab and the Interactive Design Center. Both facilities are integral to the successful achievement of mission objectives within CS & IS: to develop innovative technologies for collaborative, high-performance computing across the nuclear-weapons complex and to provide modeling, simulation, and visualization capabilities in an interactive environment that permits the analysis and design of nuclear weapons in real time.
The majority of our work is supported by the National Nuclear Security Administration’s Advanced Simulation and Computing (ASC) program. In addition, we have important projects with other federal agencies and Department of Energy offices, such as the Scientific Discovery through Advanced Computing program.
The key focus areas for our group include the following: advanced software; computer and network security; mathematics, informatics, and decision sciences; scalable computing; and visualization and scientific computing.
Advanced Software
Our application development work focuses on complex finite-element problem setup capabilities for weapon applications. We develop tools that address specific analysis steps and then integrate these tools into a seamless analysis environment. Our knowledge environments research is deployed as “collaboratories” that help the combustion- and biological-science communities collectively manage their work. In addition, we participate in enterprise-model development for homeland security and defense customers and have developed a widely used Java-based rules engine called JESS.
Projects in this area include the following:
Computer and Network Security
The increased use of computing is driving a rise in the value of information and the increased importance of information protection. This task is increasingly more difficult as communication networks and collaborative environments become more ubiquitous. In this focus area, we address the broad research, development, education, and operational support issues necessary to provide secure distributed computing and information management for ourselves and others.
Projects in this area include the following:
- NetState is a distributed intrusion-detection system that develops network traffic profiles and attempts to identify abnormal behavior patterns. NetState also provides status information about the network, version tracking, and connection tracking. Multiple NetState sniffers can be deployed — one at each entry point of the target network — to monitor network traffic, gather information about network sessions and software versions, and send this information to the NetState server for storage in a centralized database.
Mathematics, Informatics, and Decision Sciences
Sandia’s computer sciences and mathematics capabilities are also brought to bear on the analysis of large-scale data sets. These data sets, found in scientific computing and other national-security applications are often heterogeneous, erroneous, and overwhelming in the rate that they are generated. Such data sets require novel algorithms, visualization capabilities, and high-performance computing architectures to support their transformation into useful information.
Projects in this area include the following:
- The ASC Data Discovery project is developing a collection of algorithms and software to detect and characterize information in massive data sets.
- The goal of the Datapipelining for Heterogeneous Data Fusion project is to make meaningful interpretations of data types that provide different views of the same situation, give complementary information despite appearing dissimilar, and are collected and stored in a variety of formats. Specifically, we are applying various ensemble-classification techniques to the problem of protein phosphorylation prediction. We are also interested in the problems of sensor fusion and medical surveillance.
- The Distributed Design and Analysis of Computer Experiments (DDACE) package is a valuable tool for analysts and engineers who want to assess the sensitivity of the response surface (output) to individual input parameters, determine which parameters contribute most to variance, and evaluate how parameters interact with one another. Additionally, DDACE can generate parameter-dependent sensitivity information for use in optimization algorithms.
- The Ensemble Text Analysis project is developing a machine-learning metamethod known as “ensembles,” which can improve a machine’s accuracy and robustness of named entity recognition (NER), the process of assigning category labels to words. The core ensemble idea is to first generate a large number of different but related NER models and then combine them to generate better output labelings.
- OPT++ is a library of nonlinear optimization algorithms written in C++. Current research centers on the algorithmic development of parallel derivative-free and gradient-based optimization methods to solve simulation-based optimization problems.
Scalable Computing
The increased reliance on modeling and simulation for engineering design using high-performance computing environments requires the development of new computational algorithms and software systems. This is particularly true in the case of terascale and distributed computing systems, since such systems have a higher probability of failure. We research and develop new computational algorithms and component-based software systems. We then integrate these algorithms and systems into robust and reliable problem-solving environments.
Projects in this area include the following:
- Ccaffeine is a Sandia-developed framework that is compliant with common component architecture, an open standard for high-performance middleware. Ccaffeine supports high-performance components and is used in projects at Sandia, the Department of Energy, and universities and industries around the world.
- MPQC is a full-fledged, massively parallel quantum-chemistry program that uses first principles to compute properties of atoms and molecules. MPQC features an internal coordinate optimizer that efficiently optimizes molecules with many degrees of freedom and that is able to handle nearly arbitrary internal coordinate constraints. Massively parallel implementations of linear-scaling local correlation methods are currently underway.
- We are partnering with industry and other research institutions to address the high-performance computing need for a versatile storage architecture that is scalable, sharable, and secure. Our research explores parallel and remote direct-memory-access technologies for scalability, promotes a standardized file system interface to facilitate sharing, and is compatible with Sandia’s corporate security infrastructure.
Visualization and Scientific Computing
Scientists and engineers today are using high-fidelity modeling and simulation codes to perform virtual prototyping and visualization capabilities. The expected sizes of data sets generated by these approaches will require compute capacity that far exceeds currently developed and planned solutions. This focus area is aimed at the research, design, integration, and production issues needed to deliver commodity-based compute and visualization systems to meet Sandia’s present and future high-performance, scientific-computing needs.
Projects in this area include the following:
- Programmable hardware shaders, which can replace the fixed-functionality vertex and pixel shaders traditionally found on graphics processing units, allow greater control of how data are rendered to images, and make possible the use of sophisticated rendering techniques at interactive frame rates.
- Our reconfigurable computing research, which seeks to improve the capabilities of high-performance computing systems, has demonstrated the use of field-programmable gate arrays (FPGAs) as hardware accelerators for scientific-computing applications. We are working to improve performance of FPGAs by refining tools and hardware.
- OVIS, a software tool for cluster-computing systems, enables system administrators to find and resolve potentially catastrophic problems in advance by detecting aberrant node behavior. OVIS developers are working on a parallel version to address scalability issues. Bayesian networks will also be incorporated into OVIS to aid in problem diagnosis.
- FCLib, a C library toolkit of building blocks that can be used to assemble analyses for data discovery, was constructed to meet the needs of data discovery in large-scale, spatio-temporal data. FCLib supports feature-based analysis, minimizes low-level processing, is easy to use, and has applications in a wide variety of science domains.
- The Sandia Higher-Order Elements (SHOE) research program focuses on visualizing higher-order, finite-element simulation results. SHOE has developed robust analysis and visualization techniques, as well as a programming interface that accurately samples higher-order finite elements when isosurfacing.
For more information, contact Jerry Friesen at (925) 294-3144.
