The Year in Perspective



Horst D. Simon image

This will be the last time that I will write this introduction to the NERSC annual report. When you read this, it will be almost 12 years since I returned to Berkeley and took on the challenge together with many dedicated colleagues to rebuild NERSC as a new center. One of the purposes of moving NERSC to Berkeley Lab in 1996 was to bring the benefits of supercomputing to a wider array of applications than was previously supported. One of the first new additions to NERSC’s scientific roster was astrophysics, with Saul Perlmutter’s Supernova Cosmology Project and with studies of cosmic microwave background radiation (CMB) by George Smoot’s research group. Both groups had already been using smaller computers for data analysis but were eager to tackle larger datasets using NERSC’s resources.

Those new collaborations bore fruit almost immediately, with Perlmutter’s co-discovery in 1998 of the dark energy that is accelerating the expansion of the Universe, and with the BOOMERANG Consortium’s discovery in 2000, based on analysis of CMB data, that the geometry of the Universe is flat. The latter finding was based on Smoot’s earlier discovery, with John Mather and colleagues, of the blackbody form and anisotropy of the CMB, for which Smoot and Mather have been honored with the 2006 Nobel Prize for physics.

Over the past ten years, George Smoot and his colleagues have used nearly 5 million processor-hours and tens of terabytes of disk space at NERSC, and around 100 analysts from a dozen CMB experiments are now NERSC users. In fact, NERSC has become the computational center around which this community has coalesced as both its computational demands and results increase dramatically. The CMB sky map that Smoot and Mather produced in 1992 used only 6,144 pixels, but the newest CMB model can map 75 billion observations to 150 million pixels, making cosmology a highly precise science.

Research at NERSC over the past year by other astrophysicists has led to new insights into the formation of high-mass stars as well as the forces that make supernovas explode. Getting back down to earth, climate researchers have shown that the global warming that has already happened will produce bigger hurricanes, longer heat waves, and more extreme weather by the end of this century.

Materials scientists and chemists have also had a very productive year at NERSC. One research group achieved a breakthrough in surface plasmon resonance, a common but, up till now, expensive technique for measuring binding interactions, such as those between DNA and proteins. They developed a low-cost crystal array to make a portable and highly sensitive sensor that can be used in diagnostic bioassays and in research ranging from drug discovery to immunology, virology, and other fields. Another group’s calculations show that zigzag graphene nanoribbons could serve as the basis for nanosized spintronic devices. Other researchers have shown why doping strengthens grain boundaries and how to overcome nanocrystals’ resistance to doping.

Plasma physics saw the development of the first self-consistent model for the spontaneous onset of fast magnetic reconnection. This new finding may help scientists to better predict which solar storms pose the greatest threat to communications and other satellites, and it may also lead to a better understanding of how to control plasmas in fusion reactors.

An INCITE award of 2.5 million processor hours at NERSC was used to create full-scale, three-dimensional, explicit particle models that revealed important physical details of laser wakefield accelerator experiments that accelerated electron beams to energies exceeding a billion electron volts (GeV) in a distance of just 3.3 centimeters. These simulations coupled with experiments are developing the detailed understanding of laser acceleration needed to apply this technology to future higher energy particle physics experiments and to compact machines for medicine and laboratory science.

On the technical side, NERSC has taken a major step forward with the acquisition of what will be the largest Cray XT4 system in the world. When completed later in 2007, the system will have more than 19,000 processors and will deliver sustained performance of at least 16 teraflop/s when running a suite of diverse scientific applications at scale. This system will increase NERSC’s sustained computational capability by almost a factor of 10.

NERSC has also laid the foundation for an analytics infrastructure that combines hardware, software, and the development and application of analytics technologies such as data management, data analysis and data mining, visual data exploration, and workflow management. These technologies will help NERSC users spend more time doing research and less time managing data and struggling with analytics software.

When the DOE’s Advanced Scientific Computing Research Advisory Committee (ASCAC) formed a subcommittee to develop performance metrics for petascale facilities, it was natural for NERSC to take a leadership role, since we have long used goals and metrics to ensure that what we do is meeting the needs of DOE and its scientists. Among the subcommittee’s recommendations were project-specific services, like those NERSC provides to SciDAC and INCITE projects, and the use of a standard user survey based on the one NERSC has used for several years to measure and improve service.

Looking to the future, NERSC is collaborating with computer scientists to meet the software challenges of petascale computing and manycore architectures, which will have hundreds to thousands of cores per processor. The explosion in hardware parallelism will require a complete redesign of applications, libraries, and algorithms—and possibly new programming languages—to fully utilize petascale systems. We are heading into a time of great innovation in high performance computing, and as always, NERSC will be influencing and implementing those innovations.

As I look back on my decade as director of the NERSC Center Division, I feel tremendous gratitude for the dedication, skills, and accomplishments of the NERSC staff, who are never satisfied with the status quo and are always looking for ways to make NERSC an even more productive resource for scientific researchers. When I hand over leadership of the division to my successor, I will do so with the confidence that he or she will have the support of the best scientific computing organization in the world.

Horst D. Simon
NERSC Center Division Director