<< Back to November 2006 CRN Table of Contents

[Published originally in the November 2006 edition of Computing Research News, Vol. 18/No. 5]

Research With Impact: The Computer Science Research Institute at Sandia National Laboratories

By David E. Womble

The Computer Science Research Institute (CSRI) at Sandia National Laboratories brings together researchers from universities, industry and the national laboratories to conduct leading-edge research in computer science, computational science and mathematics to provide new capabilities in modeling and simulation, and to apply this capability for our nation’s security.

The CSRI is funded by DOE’s Advanced Simulation and Computing (ASC) program and is a key part of this program’s engagement with a broad research community.

Sandia and the CSRI have been leaders in computer science and computational science for many years. Since 1985, Sandia has pioneered efforts to develop scalable computing and apply this capability to problems in national security. In 1987, Sandia researchers won the first Gordon Bell Award as well as the Karp Challenge. Sandia researchers have continued to win awards, receive patents and work with industry—notably Intel and Cray—to design, develop and deploy several of the most powerful systems in the world. The latest of these machines, “Red Storm,” is now available commercially from Cray as the XT3. Sandia architecture and system software expertise are directly responsible for the success of these platforms.

Just as importantly, Sandia researchers use high-performance computing to solve problems in national security. Sandia codes and algorithms form the basis of simulations used to insure the safety and security of the nation’s nuclear weapon stockpile. Sandia researchers have also solved problems in the areas of informatics, energy, infrastructure analysis, homeland security, biology, fusion, and nanotechnology. Sandia codes and libraries are widely used by industry, academia and the national laboratories. These codes include Trilinos (solvers), DAKOTA (optimization and uncertainty quantification), Zoltan (load balancing), CUBIT (meshing) and ParaView (visualization).

The CSRI plays a critical role in this mission. CSRI researchers collaborate in areas such as computer architectures and networking, operating systems, applied mathematics, algorithms, visualization, and information science. The CSRI provides both a physical and technical focal point for identifying problems, conducting research and developing and strengthening interactions between the university and laboratory researchers.

Participants are also encouraged to develop long-term relationships with laboratory scientists and researchers. In addition to strong collaborative research programs, the CSRI sponsors:

• summer faculty and summer intern programs with up to 60 participants each year;
• faculty sabbaticals;
• four to eight workshops each year targeting areas with especially strong opportunities for impact;
• graduate student programs and research fellowships;
• a strong visitor program with both short-term and extended visits; and
• a technical seminar series.

In addition to the strong research component in CSRI activities, the CSRI encourages participating students to choose careers in computer science, computational science and mathematics that directly support the challenges of national security.

Research and Technical Focus Areas

The CSRI conducts research in computer science, computational science and mathematics that impacts large-scale modeling and simulation. One area of particular interest is developing hardware, software and algorithms that scale both as the size and complexity of the problems increase and as the number of processors increase to tens or hundreds of thousands.

The technical focus of the CSRI can be divided into two areas, algorithms and enabling technology along with advanced architectures and system software. The near-term focus in the advanced architectures and system software area is the design of 20 PFlop systems that are scalable, reliable and usable by scientists and engineers. Moore’s Law is still in effect in the semiconductor industry, and success in this area will require advances in processor architectures and design, integrated and balanced interconnects, fault tolerance, scalable I/O and file systems, scalable operating systems, and improved programming environments.

Longer term computer research at the CSRI looks beyond the realm of Moore’s Law. There is a specific focus on the use of nanotechnologies to the computing industry, including novel memory technologies, reversible logic and quantum computing. CSRI researchers work closely with Sandia physicists and materials scientists to take these new technologies beyond speculation and into reality.

The area of algorithms and enabling technologies includes research in mathematics and computer science and deals with the question of how to use high-end computing to provide a modeling and simulation capability for scientists and engineers. Linear and nonlinear solvers form the computational kernel of many, if not most, simulations, and CSRI researchers have made tremendous contributions in this area. Many of these advances are incorporated in the well-known Trilinos framework.

CSRI researchers address the fundamental questions of how modeling and simulation is used through their work in optimization and uncertainty quantification. Researchers in this area address the related areas of validation and verification and the linkage between simulation and experiment that will be an essential part of any future scientific or engineering endeavor.

Enabling technologies play a critical role in the use of high-performance modeling and simulation platforms, and CSRI researchers have developed some of the best tools available. These include the CUBIT meshing toolkit (with associated mesh optimization and geometry editing toolkits) and the ParaView visualization package. ParaView, in particular, addresses the scalability of the user environment by allowing the same visualization tools to be used both on the desktop and in high-end visualization laboratories.

Finally, information-based computing is increasingly important in many national-security applications. To address this national security need, the CSRI has continued to build a research focus in graph-based and discrete algorithms. This area has been particularly interesting because it has introduced new areas of research at the interface of algorithms, architectures and visualization as scientists begin to incorporate an “information-based” approach to their work, in addition to the more traditional “geometry-based” approach.

In addition to the foundational areas described above, the CSRI emphasizes collaborations in one or two focus areas each year. These areas are determined through a combination of the needs of national security and the interest of CSRI collaborators. In 2006, there was a special focus on simulation in nanotechnology. Researchers considered both the simulation of nano-devices and the use of nano-devices in computing. The interface between the micro and nano scales was of special interest.

In 2007, the CSRI will emphasize two areas. The first is verification and validation, particularly for simulations involving multiple scales. The second area is software environments that will be critical as petaflop computing comes online.

Facilities

The Computer Science Research Institute has just moved into new facilities specifically designed and built for collaboration. Located in Albuquerque, NM, it is easily accessible to visitors. The building comprises 34,500 square feet and includes space for 130 research staff and 50 collaborators. It includes a conference room measuring 2,000 square feet with seating for up to 80, a computer lab and four additional collaborative work areas and conference rooms. The CSRI also has space at its Livermore, California location.

CSRI researchers have access to extensive computing facilities. These facilities include the Red Storm supercomputer (over 13,000 processors, 125 TFlops), the Institutional Computing Cluster (over 4,000 processors), Thunderbird (over 4,000 processors), and over 10 clusters of various sizes up to approximately 500 processors specifically dedicated to software and algorithm development.

About Sandia National Laboratories

Since 1949, Sandia National Laboratories has developed science-based technologies that support our national security. Today, nearly 300 million Americans depend, even if unknowingly, on Sandia's technology solutions to solve national and global threats to peace and freedom. Sandia relies on strong fundamental programs in science, technology, and engineering to underpin and build critical R&D capabilities that support our national security mission areas: nuclear weapons, energy and infrastructure assurance, nonproliferation, defense systems and assessments and homeland security. Sandia employs more than 8,000 staff members; of these, over 1,500 have earned doctorates and over 2,500 have earned masters degrees in physical sciences and engineering disciplines.

Sandia is a government-owned, contractor-operated facility. Sandia Corporation, a Lockheed Martin company, manages Sandia for the U.S. Department of Energy's National Nuclear Security Administration.

(Picture: The new CSRI building in Albuquerque, NM facilitates collaborative research.)

David Womble is the Senior Manager of the Computer Science Research Institute (CSRI) and the Computer Science and Mathematics Group at Sandia National Laboratories.

Note: This is another in a series of CRN articles describing the activities of CRA’s laboratory members. Others are posted at: http://www.cra.org/reports/labs.