Aggressive technology scaling trends have worsened the transient fault problem in high-performance computing (HPC) systems. Some faults are benign, but others can lead to silent data corruption (SDC), which represents a serious problem; a fault introducing an error that is not readily detected nto an HPC simulation. Due to the insidious nature of SDCs, researchers have worked to understand their impact on applications. Previous studies have relied on expensive fault injection campaigns with uniform sampling to provide overall SDC rates, but this solution does not provide any feedback on the code regions without samples.

The trend of rapid technology scaling is expected to make the hardware of high-performance computing (HPC) systems more susceptible to computational errors due to random bit flips. Some bit flips may cause a program to crash or have a minimal effect on the output, but others may lead to silent data corruption (SDC), i.e., undetected yet significant output errors. Classical fault injection analysis methods employ uniform sampling of random bit flips during program execution to derive a statistical resiliency profile. However, summarizing such fault injection result with sufficient detail is difficult, and understanding the behavior of the fault-corrupted program is still a challenge. In this article, we introduce SpotSDC, a visualization system to facilitate the analysis of a program's resilience to SDC. SpotSDC provides multiple perspectives at various levels of detail of the impact on the output relative to where in the source code the flipped bit occurs, which bit is flipped, and when during the execution it happens. SpotSDC also enables users to study the code protection and provide new insights to understand the behavior of a fault-injected program. Based on lessons learned, we demonstrate how what we found can improve the fault injection campaign method.

Abstract Parallel simulation codes often suffer from performance bottlenecks due to network congestion, leaving millions of dollars of investments underutilized. Given a network topology, it is critical to understand how different applications, job placements, routing schemes, etc., are affected by and contribute to network congestion, especially for large and complex networks. Understanding and optimizing communication on large‐scale networks is an active area of research. Domain experts often use exploratory tools to develop both intuitive and formal metrics for network health and performance. This paper presentsTreeScope, an interactive, web‐based visualization tool for exploring network traffic on large‐scale fat‐tree networks.TreeScopeencodes the network topology using a tailored matrix‐based representation and provides detailed visualization of all traffic in the network. We report on the design process ofTreeScope, which has been received positively by network researchers as well as system administrators. Through case studies of real and simulated data, we demonstrate howTreeScope's visual design and interactive support for complex queries on network traffic can provide experts with new insights into the occurrences and causes of congestion in the network.

Constructing distributed representations for words through neural language models and using the resulting vector spaces for analysis has become a crucial component of natural language processing (NLP). However, despite their widespread application, little is known about the structure and properties of these spaces. To gain insights into the relationship between words, the NLP community has begun to adapt high-dimensional visualization techniques. In particular, researchers commonly use t-distributed stochastic neighbor embeddings (t-SNE) and principal component analysis (PCA) to create two-dimensional embeddings for assessing the overall structure and exploring linear relationships (e.g., word analogies), respectively. Unfortunately, these techniques often produce mediocre or even misleading results and cannot address domain-specific visualization challenges that are crucial for understanding semantic relationships in word embeddings. Here, we introduce new embedding techniques for visualizing semantic and syntactic analogies, and the corresponding tests to determine whether the resulting views capture salient structures. Additionally, we introduce two novel views for a comprehensive study of analogy relationships. Finally, we augment t-SNE embeddings to convey uncertainty information in order to allow a reliable interpretation. Combined, the different views address a number of domain-specific tasks difficult to solve with existing tools.

Controlling the spread of infectious diseases in large populations is an important societal challenge. Mathematically, the problem is best captured as a certain class of reaction-diffusion processes (referred to as contagion processes) over appropriate synthesized interaction networks. Agent-based models have been successfully used in the recent past to study such contagion processes. We describe EpiSimdemics, a highly scalable, parallel code written in Charm++ that uses agent-based modeling to simulate disease spreads over large, realistic, co-evolving interaction networks. We present a new parallel implementation of EpiSimdemics that achieves unprecedented strong and weak scaling on different architectures - Blue Waters, Cori and Mira. EpiSimdemics achieves five times greater speedup than the second fastest parallel code in this field. This unprecedented scaling is an important step to support the long term vision of realtime epidemic science. Finally, we demonstrate the capabilities of EpiSimdemics by simulating the spread of influenza over a realistic synthetic social contact network spanning the continental United States (~280 million nodes and 5.8 billion social contacts).

The dragonfly topology is a popular choice for building high-radix, low-diameter, hierarchical networks with high-bandwidth links. On Cray installations of the dragonfly network, job placement policies and routing inefficiencies can lead to significant network congestion for a single job and multi-job workloads. In this paper, we explore the effects of job placement, parallel workloads and network configurations on network health to develop a better understanding of inter-job interference. We have developed a functional network simulator, Damselfly, to model the network behavior of Cray Cascade, and a visual analytics tool, DragonView, to analyze the simulation output. We simulate several parallel workloads based on five representative communication patterns on up to 131,072 cores. Our simulations and visualizations provide unique insight into the buildup of network congestion and present a trade-off between deployment dollar costs and performance of the network.

Biosurveillance is a critical area in the intelligence community for real-time detection of disease outbreaks. Identifying epidemics enables analysts to detect and monitor disease outbreak that might be spread from natural causes or from possible biological warfare attacks. Containing these events and disseminating alerts requires the ability to rapidly find, classify and track harmful biological signatures. In this paper, we describe a novel visual paradigm to conduct biosurveillance using an Infectious Disease Weather Map. Our system provides a visual common ground in which users can view, explore and discover emerging concepts and correlations such as symptoms, syndromes, pathogens and geographic locations.

Advances in surveillance science have supported public health agencies in tracking and responding to disease outbreaks. Increasingly, epidemiologists have been tasked with interpreting multiple streams of heterogeneous data arising from varied surveillance systems. As a result public health personnel have experienced an overload of plots and charts as information visualization techniques have not kept pace with the rapid expansion in data availability. This study sought to advance the science of public health surveillance data visualization by conceptualizing a visual paradigm that provides an 'epidemiological canvas' for detection, monitoring, exploration and discovery of regional infectious disease activity and developing a software prototype of an 'infectious disease weather map'. Design objectives were elucidated and the conceptual model was developed using cognitive task analysis with public health epidemiologists. The software prototype was pilot tested using retrospective data from a large, regional pediatric hospital, and gastrointestinal and respiratory disease outbreaks were re-created as a proof of concept.

Radial visualization, or the practice of displaying data in a circular or elliptical pattern, is an increasingly common technique in information visualization research. In spite of its prevalence, little work has been done to study this visualization paradigm as a methodology in its own right. We provide a historical review of radial visualization, tracing it to its roots in centuries-old statistical graphics. We then identify the types of problem domains to which modern radial visualization techniques have been applied. A taxonomy for radial visualization is proposed in the form of seven design patterns encompassing nearly all recent works in this area. From an analysis of these patterns, we distill a series of design considerations that system builders can use to create new visualizations that address aspects of the design space that have not yet been explored. It is hoped that our taxonomy will provide a framework for facilitating discourse among researchers and stimulate the development of additional theories and systems involving radial visualization as a distinct design metaphor.

We propose a novel point-based approach to view dependent isosurface extraction. We introduce a fast visibility query system for the view dependent traversal, which exhibits moderate memory requirements. This technique allows for an interactive interrogation of the full visible woman dataset (1GB) at four to fifteen frames per second on a desktop computer. The point-based approach is built on an extraction scheme that classifies different sections of the isosurface into four categories, depending on the size of the geometry when projected onto the screen. In particular, we use points to represent small and subpixel triangles, as well as larger sections of the isosurface whose projection has subpixel size. To assign consistent and robust normals to individual points representing such regions, we propose to compute them during post processing of the extracted isosurface and provide the corresponding hardware implementation.

Presents a brute-force ray-tracing system for interactive volume visualization. The system runs on a conventional (distributed) shared-memory multiprocessor machine. For each pixel, we trace a ray through a volume to compute the color for that pixel. Although this method has a high intrinsic computational cost, its simplicity and scalability make it ideal for large data sets on current high-end parallel systems. To gain efficiency, several optimizations are used, including a volume bricking scheme and a shallow data hierarchy. These optimizations are used in three separate visualization algorithms: isosurfacing of rectilinear data, isosurfacing of unstructured data, and maximum-intensity projection on rectilinear data. The system runs interactively (i.e. at several frames per second) on an SGI Reality Monster. The graphics capabilities of the Reality Monster are used only for display of the final color image.

We propose a new approach to polygonal isosurface extraction that is based on extracting only the visible portion of the isosurface. The visibility tests are done in two phases. First, coarse visibility tests are performed in software to determine the visible cells. These tests are based on hierarchical tiles and shear-warp factorization. The second phase resolves the visible portions of the extracted triangles and is accomplished by the graphics hardware. While the latest isosurface extraction methods have effectively eliminated the search phase bottleneck, the cost of constructing and rendering the isosurface remains high. Many of today's large datasets contain very large and complex isosurfaces that can easily overwhelm even state-of-the-art graphics hardware. The proposed approach is output sensitive and is thus well suited for remote visualization applications where the extraction and rendering phases are done on a separate machines.

We show that it is feasible to perform interactive isosurfacing of very large rectilinear datasets with brute-force ray tracing on a conventional (distributed) shared-memory multiprocessor machine. Rather than generate geometry representing the isosurface and render with a z-buffer, for each pixel we trace a ray through a volume and do an analytic isosurface intersection computation. Although this method has a high intrinsic computational cost, its simplicity and scalability make it ideal for large datasets on current high-end systems. Incorporating simple optimizations, such as volume bricking and a shallow hierarchy, enables interactive rendering (i.e. 10 frames per second) of the 1 GByte full resolution Visible Woman dataset on an SGI Reality Monster. The graphics capabilities of the Reality Monster are used only for display of the final color image.

The recent years have seen rapid advancement towards viewing the Earth as an integrated system. This means that we have come to understand the interdependence of the major planetary subsystems-atmosphere, biosphere, oceans and the deep earth interior-on a large range of time and length scales. One of the longest time scales of the planet is imposed by solid state convection within the silicate Earth mantle. Mantle convection modeling, and other earth science modeling efforts, now are producing simulation data on grids that are large enough to strain the memory and processing power of even the largest high-end graphics workstations. Another alternative is to use parallel visualization tools running on the massively parallel computers that generated the data. This is the approach that we have taken for the visualization of mantle convection simulation data.

We present efficient sequential and parallel algorithms for isosurface extraction. Based on the Span Space data representation, new data subdivision and searching methods are described. We also present a parallel implementation with an emphasis on load balancing. The performance of our sequential algorithm to locate the cell elements intersected by isosurfaces is faster than the Kd tree searching method originally used for the Span Space algorithm. The parallel algorithm can achieve high load balancing for massively parallel machines with distributed memory architectures.

Presents the "Near Optimal IsoSurface Extraction" (NOISE) algorithm for rapidly extracting isosurfaces from structured and unstructured grids. Using the span space, a new representation of the underlying domain, we develop an isosurface extraction algorithm with a worst case complexity of o(/spl radic/n+k) for the search phase, where n is the size of the data set and k is the number of cells intersected by the isosurface. The memory requirement is kept at O(n) while the preprocessing step is O(n log n). We utilize the span space representation as a tool for comparing isosurface extraction methods on structured and unstructured grids. We also present a fast triangulation scheme for generating and displaying unstructured tetrahedral grids.