PNNL has deep expertise in energy-efficient computing; performance, power, and reliability modeling; exploration and design of novel computing architectures; and data-driven discovery at extreme scales. Specific domain areas include architectures, programming models, resiliency, fault tolerance, information visualization, data analytics, and data management.
Our work programming models for extreme-scale computing includes toolkits such as Global Arrays, which powers NWChem and other scientific applications, including the subsurface flow modeling code STOMP and power grid modeling code GridPACKTM. PNNL data scientists lead research in data exploitation, workflow, and provenance at extreme scales for science, energy, and security domains.
In the field of visual analytics and exploratory data analysis, we have developed visual metaphors for complex, high-volume data and signature discovery algorithms that apply advanced statistics and machine learning to derive novel indicators of complex phenomena. Recent investments have focused on using PNNL’s visualization capabilities to aid in interpretability and interactive tuning of machine learning algorithms. PNNL is also making significant advances in graph analytics, including hybrid architectures for exploiting large graph data sets and algorithms for graph query on multi-threaded systems.
Resources include the Performance and Architecture Laboratory. This lab is used for measuring performance, power, and thermal effects on advanced technologies to predict their overall potential and guide their designs. Additional computing resources include the 3.4 petaflop Cascade supercomputer, Research Computing mid-range cluster, Marianas GPU cluster, and a research cloud. In addition, the Center for Advanced Technology Evaluation (CENATE) offers a testbed for computing technologies. Finally, specialized laboratories include those for human-computer interaction research for visual interfaces, and emerging virtual reality environments.
These capabilities receive support through programs from the Department of Energy’s Applied Scientific Computing Research, Basic Energy Sciences, Biological and Environmental Research, High Energy Physics, Energy Efficiency and Renewable Energy, Fossil Energy, National Nuclear Security Administration, Department of Homeland Security, and other sponsors, including the Department of Health and Human Services and the Department of Defense.