Engineering topics

Ellsworth, David

Publications and source records attributed to Ellsworth, David.

9 records

NASA NTRS2019

Electra: A Modular-Based Expansion of NASA's Supercomputing Capability

NASA has increasingly relied on high-performance computing (HPC) re- sources for computational modeling, simulation, and data analysis to meet the science and engineering goals of its missions in space exploration, aeronautics, and Earth and space science. The NASA Advanced Supercomputing (NAS) Division at Ames Research Center in Silicon Valley, Calif., hosts NASA’s premier supercomputing resources, integral to achieving and enhancing the success of the agency’s missions. NAS provides a balanced environment, funded under the High-End Computing Capability (HECC) project, comprised of world-class supercomputers, including its flagship distributed-memory cluster, Pleiades; high-speed networking; and massive data storage facilities, along with multi-disciplinary support teams for user support, code porting and optimization, and large-scale data analysis and scientific visualization. However, as scientists have increased the fidelity of their simulations and engineers are conducting larger parameter-space studies, the requirements for supercomputing resources have been growing by leaps and bounds. With the facility housing the HECC systems reaching its power and cooling capacity, NAS undertook a prototype project to investigate an alternative approach for housing supercomputers. Modular supercomputing, or container-based computing, is an innovative concept for expanding NASA’s HPC capabilities. With modular supercomputing, additional containers—similar to portable storage pods—can be connected together as needed to accommodate the agency’s ever-increasing demand for computing resources. In addition, taking advantage of the local weather permits the use of cooling technologies that would additionally save energy and reduce annual water usage. The first stage of NASA’s Modular Supercomputing Facility (MSF) prototype, which resulted in a 1,000 square-foot module on a concrete pad with room for 16 compute racks, was completed in Fall 2016 and an SGI (now HPE) computer system, named Electra, was deployed there in early 2017. Cooling is performed via an evaporative system built into the module, and preliminary experience shows a Power Usage Effectiveness (PUE) measurement of 1.03. Electra achieved over a petaflop on the LINPACK benchmark, sufficient to rank number 96 on the November 2016 TOP500 list [14]. The system consists of 1,152 InfiniBand-connected Intel Xeon Broadwell-based nodes. Its users access their files on a facility-wide file system shared by all HECC compute assets via Mellanox MetroX InfiniBand extenders, which connect the Electra fabric to Lustre routers in the primary facility over fiber-optic links about 900 feet long. The MSF prototype has exceeded expectations and is serving as a blueprint for future expansions. In the remainder of this chapter, we detail how modular data center technology can be used to expand an existing compute resource. We begin by describing NASA’s requirements for supercomputing and how resources were provided prior to the integration of the Electra module-based system.

Biswas, Rupak↗

NASA NTRS2006

Multiple-Flat-Panel System Displays Multidimensional Data

The NASA Ames hyperwall is a display system designed to facilitate the visualization of sets of multivariate and multidimensional data like those generated in complex engineering and scientific computations. The hyperwall includes a 77 matrix of computer-driven flat-panel video display units, each presenting an image of 1,280 1,024 pixels. The term hyperwall reflects the fact that this system is a more capable successor to prior computer-driven multiple-flat-panel display systems known by names that include the generic term powerwall and the trade names PowerWall and Powerwall. Each of the 49 flat-panel displays is driven by a rack-mounted, dual-central-processing- unit, workstation-class personal computer equipped with a hig-hperformance graphical-display circuit card and with a hard-disk drive having a storage capacity of 100 GB. Each such computer is a slave node in a master/ slave computing/data-communication system (see Figure 1). The computer that acts as the master node is similar to the slave-node computers, except that it runs the master portion of the system software and is equipped with a keyboard and mouse for control by a human operator. The system utilizes commercially available master/slave software along with custom software that enables the human controller to interact simultaneously with any number of selected slave nodes. In a powerwall, a single rendering task is spread across multiple processors and then the multiple outputs are tiled into one seamless super-display. It must be noted that the hyperwall concept subsumes the powerwall concept in that a single scene could be rendered as a mosaic image on the hyperwall. However, the hyperwall offers a wider set of capabilities to serve a different purpose: The hyperwall concept is one of (1) simultaneously displaying multiple different but related images, and (2) providing means for composing and controlling such sets of images. In place of elaborate software or hardware crossbar switches, the hyperwall concept substitutes reliance on the human visual system for integration, synthesis, and discrimination of patterns in complex and high-dimensional data spaces represented by the multiple displayed images. The variety of multidimensional data sets that can be displayed on the hyperwall is practically unlimited. For example, Figure 2 shows a hyperwall display of surface pressures and streamlines from a computational simulation of airflow about an aerospacecraft at various Mach numbers and angles of attack. In this display, Mach numbers increase from left to right and angles of attack increase from bottom to top. That is, all images in the same column represent simulations at the same Mach number, while all images in the same row represent simulations at the same angle of attack. The same viewing transformations and the same mapping from surface pressure to colors were used in generating all the images.

Gundo, Daniel↗

NASA NTRS2004

Interactive Terascale Particle Visualization

This paper describes the methods used to produce an interactive visualization of a 2 TB computational fluid dynamics (CFD) data set using particle tracing (streaklines). We use the method introduced by Bruckschen et al. [2001] that pre-computes a large number of particles, stores them on disk using a space-filling curve ordering that minimizes seeks, and then retrieves and displays the particles according to the user's command. We describe how the particle computation can be performed using a PC cluster, how the algorithm can be adapted to work with a multi-block curvilinear mesh, and how the out-of-core visualization can be scaled to 296 billion particles while still achieving interactive performance on PG hardware. Compared to the earlier work, our data set size and total number of particles are an order of magnitude larger. We also describe a new compression technique that allows the lossless compression of the particles by 41% and speeds the particle retrieval by about 30%.

Ellsworth, David↗

NASA NTRS2003

Accelerating Large Data Analysis By Exploiting Regularities

We present techniques for discovering and exploiting regularity in large curvilinear data sets. The data can be based on a single mesh or a mesh composed of multiple submeshes (also known as zones). Multi-zone data are typical to Computational Fluid Dynamics (CFD) simulations. Regularities include axis-aligned rectilinear and cylindrical meshes as well as cases where one zone is equivalent to a rigid-body transformation of another. Our algorithms can also discover rigid-body motion of meshes in time-series data. Next, we describe a data model where we can utilize the results from the discovery process in order to accelerate large data visualizations. Where possible, we replace general curvilinear zones with rectilinear or cylindrical zones. In rigid-body motion cases we replace a time-series of meshes with a transformed mesh object where a reference mesh is dynamically transformed based on a given time value in order to satisfy geometry requests, on demand. The data model enables us to make these substitutions and dynamic transformations transparently with respect to the visualization algorithms. We present results with large data sets where we combine our mesh replacement and transformation techniques with out-of-core paging in order to achieve significant speed-ups in analysis.

Moran, Patrick J.↗

NASA NTRS2001

Accelerating Demand Paging for Local and Remote Out-of-Core Visualization

This paper describes a new algorithm that improves the performance of application-controlled demand paging for the out-of-core visualization of data sets that are on either local disks or disks on remote servers. The performance improvements come from better overlapping the computation with the page reading process, and by performing multiple page reads in parallel. The new algorithm can be applied to many different visualization algorithms since application-controlled demand paging is not specific to any visualization algorithm. The paper includes measurements that show that the new multi-threaded paging algorithm decreases the time needed to compute visualizations by one third when using one processor and reading data from local disk. The time needed when using one processor and reading data from remote disk decreased by up to 60%. Visualization runs using data from remote disk ran about as fast as ones using data from local disk because the remote runs were able to make use of the remote server's high performance disk array.

Ellsworth, David↗

NASA NTRS2001

Hiding the Disk and Network Latency of Out-of-Core Visualization

This paper describes an algorithm that improves the performance of application-controlled demand paging for out-of-core visualization by hiding the latency of reading data from both local disks or disks on remote servers. The performance improvements come from better overlapping the computation with the page reading process, and by performing multiple page reads in parallel. The paper includes measurements that show that the new multithreaded paging algorithm decreases the time needed to compute visualizations by one third when using one processor and reading data from local disk. The time needed when using one processor and reading data from remote disk decreased by two thirds. Visualization runs using data from remote disk actually ran faster than ones using data from local disk because the remote runs were able to make use of the remote server's high performance disk array.

Ellsworth, David↗

NASA NTRS2000

Accelerating Time-Varying Hardware Volume Rendering Using TSP Trees and Color-Based Error Metrics

This paper describes a new hardware volume rendering algorithm for time-varying data. The algorithm uses the Time-Space Partitioning (TSP) tree data structure to identify regions within the data that have spatial or temporal coherence. By using this coherence, the rendering algorithm can improve performance when the volume data is larger than the texture memory capacity by decreasing the amount of textures required. This coherence can also allow improved speed by appropriately rendering flat-shaded polygons instead of textured polygons, and by not rendering transparent regions. To reduce the polygonization overhead caused by the use of the hierarchical data structure, we introduce an optimization method using polygon templates. The paper also introduces new color-based error metrics, which more accurately identify coherent regions compared to the earlier scalar-based metrics. By showing experimental results from runs using different data sets and error metrics, we demonstrate that the new methods give substantial improvements in volume rendering performance.

Ellsworth, David↗

NASA NTRS1999

Field Encapsulation Library The FEL 2.2 User Guide

This document describes version 2.2 of the Field Encapsulation Library (FEL), a library of mesh and field classes. FEL is a library for programmers - it is a "building block" enabling the rapid development of applications by a user. Since FEL is a library intended for code development, it is essential that enough technical detail be provided so that one can make full use of the code. Providing such detail requires some assumptions with respect to the reader's familiarity with the library implementation language, C++, particularly C++ with templates. We have done our best to make the explanations accessible to those who may not be completely C++ literate. Nevertheless, familiarity with the language will certainly help one's understanding of how and why things work the way they do. One consolation is that the level of understanding essential for using the library is significantly less than the level that one should have in order to modify or extend the library. One more remark on C++ templates: Templates have been a source of both joy and frustration for us. The frustration stems from the lack of mature or complete implementations that one has to work with. Template problems rear their ugly head particularly when porting. When porting C code, successfully compiling to a set of object files typically means that one is almost done. With templated C++ and the current state of the compilers and linkers, generating the object files is often only the beginning of the fun. On the other hand, templates are quite powerful. Used judiciously, templates enable more succinct designs and more efficient code. Templates also help with code maintenance. Designers can avoid creating objects that are the same in many respects, but not exactly the same. For example, FEL fields are templated by node type, thus the code for scalar fields and vector fields is shared. Furthermore, node type templating allows the library user to instantiate fields with data types not provided by the FEL authors. This type of flexibility would be difficult to offer without the support of the language. For users who may be having template-related problems, we offer the consolation that support for C++ templates is destined to improve with time. Efforts such as the Standard Template Library (STL) will inevitably drive vendors to provide more thorough, optimized tools for template code development. Furthermore, the benefits will become harder to resist for those who currently subscribe to the least-common-denominator "code it all in C" strategy. May FEL bring you both increased productivity and aesthetic satisfaction.

Moran, Patrick J.↗

NASA NTRS1997

Application-Controlled Demand Paging for Out-of-Core Visualization

In the area of scientific visualization, input data sets are often very large. In visualization of Computational Fluid Dynamics (CFD) in particular, input data sets today can surpass 100 Gbytes, and are expected to scale with the ability of supercomputers to generate them. Some visualization tools already partition large data sets into segments, and load appropriate segments as they are needed. However, this does not remove the problem for two reasons: 1) there are data sets for which even the individual segments are too large for the largest graphics workstations, 2) many practitioners do not have access to workstations with the memory capacity required to load even a segment, especially since the state-of-the-art visualization tools tend to be developed by researchers with much more powerful machines. When the size of the data that must be accessed is larger than the size of memory, some form of virtual memory is simply required. This may be by segmentation, paging, or by paged segments. In this paper we demonstrate that complete reliance on operating system virtual memory for out-of-core visualization leads to poor performance. We then describe a paged segment system that we have implemented, and explore the principles of memory management that can be employed by the application for out-of-core visualization. We show that application control over some of these can significantly improve performance. We show that sparse traversal can be exploited by loading only those data actually required. We show also that application control over data loading can be exploited by 1) loading data from alternative storage format (in particular 3-dimensional data stored in sub-cubes), 2) controlling the page size. Both of these techniques effectively reduce the total memory required by visualization at run-time. We also describe experiments we have done on remote out-of-core visualization (when pages are read by demand from remote disk) whose results are promising.

Cox, Michael↗