Publications

Abstract not provided.

Abstract not provided.

Extreme-scale parallel systems will require alternative methods for applications to maintain current levels of uninterrupted execution. Redundant computation is one approach to consider, if the benefits of increased resiliency outweigh the cost of consuming additional resources. We describe a transparent redundancy approach for MPI applications and detail two different implementations that provide the ability to tolerate a range of failure scenarios, including loss of application processes and connectivity.We compare these two approaches and show performance results from micro-benchmarks that bound worst-case message passing performance degradation.We propose several enhancements that could lower the overhead of providing resiliency through redundancy.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Arctic sea ice plays an important role in global climate by reflecting solar radiation and insulating the ocean from the atmosphere. Due to feedback effects, the Arctic sea ice cover is changing rapidly. To accurately model this change, high-resolution calculations must incorporate: (1) annual cycle of growth and melt due to radiative forcing; (2) mechanical deformation due to surface winds, ocean currents and Coriolis forces; and (3) localized effects of leads and ridges. We have demonstrated a new mathematical algorithm for solving the sea ice governing equations using the material-point method with an elastic-decohesive constitutive model. An initial comparison with the LANL CICE code indicates that the ice edge is sharper using Materials-Point Method (MPM), but that many of the overall features are similar.

Abstract not provided.

Abstract not provided.

Abstract not provided.

The problem of missing data is ubiquitous in domains such as biomedical signal processing, network traffic analysis, bibliometrics, social network analysis, chemometrics, computer vision, and communication networks|all domains in which data collection is subject to occasional errors. Moreover, these data sets can be quite large and have more than two axes of variation, e.g., sender, receiver, time. Many applications in those domains aim to capture the underlying latent structure of the data; in other words, they need to factorize data sets with missing entries. If we cannot address the problem of missing data, many important data sets will be discarded or improperly analyzed. Therefore, we need a robust and scalable approach for factorizing multi-way arrays (i.e., tensors) in the presence of missing data. We focus on one of the most well-known tensor factorizations, CANDECOMP/PARAFAC (CP), and formulate the CP model as a weighted least squares problem that models only the known entries. We develop an algorithm called CP-WOPT (CP Weighted OPTimization) using a first-order optimization approach to solve the weighted least squares problem. Based on extensive numerical experiments, our algorithm is shown to successfully factor tensors with noise and up to 70% missing data. Moreover, our approach is significantly faster than the leading alternative and scales to larger problems. To show the real-world usefulness of CP-WOPT, we illustrate its applicability on a novel EEG (electroencephalogram) application where missing data is frequently encountered due to disconnections of electrodes.

We examine several conducting spheres moving through a magnetic field gradient. An analytical approximation is derived and an experiment is conducted to verify the analytical solution. The experiment is simulated as well to produce a numerical result. Both the low and high magnetic Reynolds number regimes are studied. Deformation of the sphere is noted in the high Reynolds number case. It is suggested that this deformation effect could be useful for designing or enhancing present protection systems against space debris.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

This talk discusses the unique demands that informatics applications, particularly graph-theoretic applications, place on computer systems. These applications tend to pose significant data movement challenges for conventional systems. Worse, underlying technology trends are moving computers to cost-driven optimization points that exacerbate the problem. The X-caliber architecture is an economically viable counter-example to conventional architectures based on the integration of innovative technologies that support the data movement requirements of large-scale informatics applications. This talk will discuss the technology drivers and architectural features of the platform, and present analysis showing the benefits for informatics applications, as well as our traditional science and engineering HPC applications.

Abstract not provided.

There is considerable interest in achieving a 1000 fold increase in supercomputing power in the next decade, but the challenges are formidable. In this paper, the authors discuss some of the driving science and security applications that require Exascale computing (a million, trillion operations per second). Key architectural challenges include power, memory, interconnection networks and resilience. The paper summarizes ongoing research aimed at overcoming these hurdles. Topics of interest are architecture aware and scalable algorithms, system simulation, 3D integration, new approaches to system-directed resilience and new benchmarks. Although significant progress is being made, a broader international program is needed.

The image created in reflected light DIC can often be interpreted as a true three-dimensional representation of the surface geometry, provided a clear distinction can be realized between raised and lowered regions in the specimen. It may be helpful if our definition of saliency embraces work on the human visual system (HVS) as well as the more abstract work on saliency, as it is certain that understanding by humans will always stand between recording of a useful signal from all manner of sensors and so-called actionable intelligence. A DARPA/DSO program lays down this requirement in a current program (Kruse 2010): The vision for the Neurotechnology for Intelligence Analysts (NIA) Program is to revolutionize the way that analysts handle intelligence imagery, increasing both the throughput of imagery to the analyst and overall accuracy of the assessments. Current computer-based target detection capabilities cannot process vast volumes of imagery with the speed, flexibility, and precision of the human visual system.

The objective of this project is to investigate the complex fracture of ice and understand its role within larger ice sheet simulations and global climate change. At the present time, ice fracture is not explicitly considered within ice sheet models due in part to large computational costs associated with the accurate modeling of this complex phenomena. However, fracture not only plays an extremely important role in regional behavior but also influences ice dynamics over much larger zones in ways that are currently not well understood. Dramatic illustrations of fracture-induced phenomena most notably include the recent collapse of ice shelves in Antarctica (e.g. partial collapse of the Wilkins shelf in March of 2008 and the diminishing extent of the Larsen B shelf from 1998 to 2002). Other fracture examples include ice calving (fracture of icebergs) which is presently approximated in simplistic ways within ice sheet models, and the draining of supraglacial lakes through a complex network of cracks, a so called ice sheet plumbing system, that is believed to cause accelerated ice sheet flows due essentially to lubrication of the contact surface with the ground. These dramatic changes are emblematic of the ongoing change in the Earth's polar regions and highlight the important role of fracturing ice. To model ice fracture, a simulation capability will be designed centered around extended finite elements and solved by specialized multigrid methods on parallel computers. In addition, appropriate dynamic load balancing techniques will be employed to ensure an approximate equal amount of work for each processor.

Abstract not provided.

Abstract not provided.

Abstract not provided.