Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Distributed memory”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 739 records · Page 41
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

BCSR on GPU: A Way Forward Extreme-scale Graph Processing on Accelerator-enabled Frontier Supercomputer

Handling large graphs in a distributed environment requires effective partitioning across processors and efficient management of local partitions. In 2D partitioning, local graphs often become too sparse, making memory-efficient data structures crucial. Using the Compressed Sparse Row (CSR) format wastes space, especially for > 83% of vertices with empty edges for the sparse graphs. This study explores bit-CSR (BCSR), a modified CSR representation, on GPUs to reduce memory usage in graph computations. We achieved 16.67% memory savings on a sparse rmat dataset with 268 million vertices and 357 million edges, without performance degradation, supported by both theoretical and experimental storage savings of 33%. However, we observed a 1.7× slowdown in degree lookup times due to bitwise operations on AMD CPUs. This analysis highlights the potential of BCSR on GPUs for improving Graph500 benchmark performance on GPU-accelerated systems, such as the Frontier supercomputer.

Sattar, Naw Safrin↗
DOE OSTI2021

Propagation Pattern for Moment Representation of the Lattice Boltzmann Method

A propagation pattern for the moment representation of the regularized lattice Boltzmann method (LBM) in three dimensions is presented. Using effectively lossless compression, the simulation state is stored as a set of moments of the lattice Boltzmann distribution function, instead of the distribution function itself. An efficient cache-aware propagation pattern for this moment representation has the effect of substantially reducing both the storage and memory bandwidth required for LBM simulations. This article extends recent work with the moment representation by expanding the performance analysis on central processing unit (CPU) architectures, considering how boundary conditions are implemented, and demonstrating the effectiveness of the moment representation on a graphics processing unit (GPU) architecture.

42 ENGINEERING↗
NASA NTRS1992

The changing nature of spacecraft operations: From the Vikings of the 1970's to the great observatories of the 1990's and beyond

Four trends in spacecraft flight operations are discussed which will reduce overall program costs. These trends are the use of high-speed, highly reliable data communications systems for distributing operations functions to more convenient and cost-effective sites; the improved capability for remote operation of sensors; a continued rapid increase in memory and processing speed of flight qualified computer chips; and increasingly capable ground-based hardware and software systems, notably those augmented by artificial intelligence functions. Changes reflected by these trends are reviewed starting from the NASA Viking missions of the early 70s, when mission control was conducted at one location using expensive and cumbersome mainframe computers and communications equipment. In the 1980s, powerful desktop computers and modems enabled the Magellan project team to operate the spacecraft remotely. In the 1990s, the Hubble Space Telescope project uses multiple color screens and automated sequencing software on small computers. Given a projection of current capabilities, future control centers will be even more cost-effective.

Ledbetter, Kenneth W.↗
NASA NTRS2016

Implementing Connected Component Labeling as a User Defined Operator for SciDB

We have implemented a flexible User Defined Operator (UDO) for labeling connected components of a binary mask expressed as an array in SciDB, a parallel distributed database management system based on the array data model. This UDO is able to process very large multidimensional arrays by exploiting SciDB's memory management mechanism that efficiently manipulates arrays whose memory requirements far exceed available physical memory. The UDO takes as primary inputs a binary mask array and a binary stencil array that specifies the connectivity of a given cell to its neighbors. The UDO returns an array of the same shape as the input mask array with each foreground cell containing the label of the component it belongs to. By default, dimensions are treated as non-periodic, but the UDO also accepts optional input parameters to specify periodicity in any of the array dimensions. The UDO requires four stages to completely label connected components. In the first stage, labels are computed for each subarray or chunk of the mask array in parallel across SciDB instances using the weighted quick union (WQU) with half-path compression algorithm. In the second stage, labels around chunk boundaries from the first stage are stored in a temporary SciDB array that is then replicated across all SciDB instances. Equivalences are resolved by again applying the WQU algorithm to these boundary labels. In the third stage, relabeling is done for each chunk using the resolved equivalences. In the fourth stage, the resolved labels, which so far are "flattened" coordinates of the original binary mask array, are renamed with sequential integers for legibility. The UDO is demonstrated on a 3-D mask of O(1011) elements, with O(108) foreground cells and O(106) connected components. The operator completes in 19 minutes using 84 SciDB instances.

UDO↗
NASA NTRSDate not supplied

Advanced eLectrical Bus (ALBus) CubeSat: From Build to Flight

The Advanced eLectrical Bus (ALBus) CubeSat project is a technology demonstration mission of a 3-U CubeSat with an advanced, digitally controlled electrical power system capability and novel use of Shape Memory Alloy (SMA) technology for reliable deployable solar array mechanisms. The objective of the project was to advance the state of power management and distribution (PMAD) capabilities to enable future missions requiring flexible and reliable power systems with higher output power capabilities. The goals of the mission were to demonstrate 100W distribution to a target electrical load, respond to continuous and fast transient power requirements, and show reliable deployment of solar arrays and antennas utilizing re-settable SMA mechanisms. The power distribution function of the ALBus PMAD system is unique in the total power to target load capability. Power is distributed from batteries to provide 100W of power directly to a resistive load. The deployable solar arrays utilize NASA Glenn Research Center’s Nickel-Titanium-Palladium-Platinum high-temperature SMAs for the retention and release mechanism, and a super-elastic binary NiTi alloy for the hinge component. The deployable solar array hinge design features the utilization of SMA material properties for a dual purpose. First, the hinge uses the super-elastic SMA properties to provide the spring force to deploy the arrays. Second, the electrical conductivity properties of the SMA enables the design to provide clean conduits for power transfer from the deployable arrays to the power management system. This eliminates the need for electrical harnesses between the arrays and the PMAD system in the ALBus system design. The uniqueness of the SMA retention and release mechanism design is the ability to reset the mechanism and reuse the same hardware. This allow functional tests of the mechanisms before flight with no degradation of performance and flying the hardware that is tested. The project launched as part of the CubeSat Launch Initiative Educational Launch of Nanosatellites (ELaNa) XIX mission on Rocket Lab’s Electron in December 2018. This paper summarizes the final launched design of ALBus and lessons learned from the technology and overall project from development through flight.

Deboshri Sadhukhan↗
NASA NTRS2018

Influence of Thin-Film Adhesives in Pullout Tests Between Nickel-Titanium Shape Memory Alloy and Carbon Fiber-Reinforced Polymer Matrix Composites

Strips of nickel-titanium (NiTi) shape memory alloy (SMA) and carbon fiber-reinforced polymer matrix composite (PMC) were bonded together using multiple thin film adhesives and their mechanical strengths were evaluated under pullout test configuration. Tensile and lap shear tests were conducted to confirm the deformation of SMAs at room temperature and to evaluate the adhesive strength between the NiTi strips and the PMC. Optical and scanning electron microscopy techniques were used to examine the interfacial bonding after failure. Simple equations on composite tensile elongation were used to fit the experimental data on tensile properties. ABAQUS models were generated to show the effects of enhanced bond strength and the distribution of stress in SMA and PMC. The results revealed that the addition of thin film adhesives increased the average adhesive strength between SMA and PMC while halting the room temperature shape memory effect within the pullout specimen.

Polymer Matrix Composites↗
NASA NTRS1980

Optical mass memory investigation

The MASTER 1 optical mass storage system advanced working model (AWM) was designed to demonstrate recording and playback of imagery data and to enable quantitative data to be derived as to the statistical distribution of raw errors experienced through the system. The AWM consists of two subsystems, the recorder and storage and retrieval. The recorder subsystem utilizes key technologies such as an acoustic travelling wave lens to achieve recording of digital data on fiche at a rate of 30 Mbits/sec, whereas the storage and retrieval reproducer subsystem utilizes a less complex optical system that employs an acousto-optical beam deflector to achieve data readout at a 5 Mbits/sec rate. The system has the built in capability for detecting and collecting error statistics. The recorder and storage and retrieval subsystems operate independent of one another and are each constructed in modular form with each module performing independent functions. The operation of each module and its interface to other modules is controlled by one controller for both subsystems.

Source record↗
DOE OSTI2025

LLM-Based Adaptive Distribution Voltage Regulation Under Frequent Topology Changes: An In-Context MPC Framework

This paper proposes a large language model (LLM) based adaptive inverter control for distribution voltage regulation under frequent topology changes. We leverage the ability of the LLM to perform in-context learning and create a topology-adaptive surrogate model for power flow calculation. The surrogate model is then integrated with a long short-term memory-based load forecaster and a model predictive control (MPC) scheme to achieve the optimal inverter control that adapts to frequent topology changes. Unlike many existing works that assume fixed-topology grids or require the knowledge of all possible topologies when training a model, the proposed in-context MPC method tackles the distribution voltage control problem under various topologies and adapts to unknown topologies with limited data requirement for fine-tuning. The effectiveness of our method is demonstrated on a modified IEEE 123-bus test system.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2026

DASSH-F: Subchannel Based Thermal Analysis

The DASSH thermal analysis code is designed to rapidly allow a reactor design engineer to obtain flow rates requirements that satisfy peak temperature constraints in the domain. The advantage of using DASSH over a hand calculation is that it has a more rigorous treatment of the pin power distribution and coolant heat transfer within an assembly and between assemblies. The advantage of using DASSH over a conventional 3D subchannel code or a computational fluid dynamics code (CFD) is that it can obtain the desired solution in a matter of minutes in serial with minor computer memory needs. The DASSH methodology for pin lattice models is virtually identical to SUPERENERGY-2 with additional functionalities taken from follow on work to SUPERENERGY-2 done at ANL in the 1980s. DASSH today is an integral component of the Argonne Fast Reactor analysis suite for reactor design work. DASSH obtains the power distribution from a coupled neutron-gamma heating calculation in GAMSOR (including DIF3D) at each time point of a companion fuel cycle analysis calculation with REBUS. The domain in DASSH assumes a hexagonal grid typical for fast reactors with much of the geometry information taken from the DIF3D model. DASSH assumes the assemblies that are loaded into each grid position are ducted to control the coolant flow. The user can alternatively provide their own geometry and power profile instead of inheriting it from DIF3D. Considerable detail is given on the subchannel formulation of DASSH in this document. Much of the formulation and design of the code builds upon research done by previous authors with little new investigation. Thus the decisions made in developing the subchannel model used in DASSH have their origins over 50 years ago. Much of the heat transfer methodology in DASSH is built upon correlations for both the coolant mixing and heat transfer coefficients for pins and ducts. DASSH is thus not a rigorous treatment of a given problem, but a rapid assessment of the temperature field that has known limitations with respect to an experimental measurement or CFD calculation. The DASSH input and output are detailed along with usage of the software. The DASSH output provides tables of evaluated material properties and key coolant and pin temperature results. DASSH can create Python scripts that generate domain summary pictures. DASSH can also generate assembly temperature maps and VTK output files which allow the DASSH solution to be visualized. As the primary purpose of the DASSH software is to compute the coolant and fuel pin temperature distribution for a given model of a reactor, much of the output focus is giving the user quick summary tables needed to assess the performance of a given orifice flow specification. The present version of DASSH has a crude orifice search capability and a sufficient orifice flow search capability. The flow search tries to meet user specified constraints for 1) peak 2-sigma clad temperature, 2) peak coolant temperature, and 3) desired bulk outlet temperature. This document serves as the manual for the Fortran based DASSH software that was developed to replace the Python version of DASSH developed as part of the VTR program.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Addressing GPU memory limitations for Graph Neural Networks in High-Energy Physics applications

Introduction Reconstructing low-level particle tracks in neutrino physics can address some of the most fundamental questions about the universe. However, processing petabytes of raw data using deep learning techniques poses a challenging problem in the field of High Energy Physics (HEP). In the Exa.TrkX Project, an illustrative HEP application, preprocessed simulation data is fed into a state-of-art Graph Neural Network (GNN) model, accelerated by GPUs. However, limited GPU memory often leads to Out-of-Memory (OOM) exceptions during training, due to the large size of models and datasets. This problem is exacerbated when deploying models on High-Performance Computing (HPC) systems designed for large-scale applications. Methods We observe a high workload imbalance issue during GNN model training caused by the irregular sizes of input graph samples in HEP datasets, contributing to OOM exceptions. We aim to scale GNNs on HPC systems, by prioritizing workload balance in graph inputs while maintaining model accuracy. Our paper introduces diverse balancing strategies aimed at decreasing the maximum GPU memory footprint and avoiding the OOM exception, across various datasets. Results Our experiments showcase memory reduction of up to 32.14% compared to the baseline. We also demonstrate the proposed strategies can avoid OOM in application. Additionally, we create a distributed multi-GPU implementation using these samplers to demonstrate the scalability of these techniques on the HEP dataset. Discussion By assessing the performance of these strategies as data loading samplers across multiple datasets, we can gauge their effectiveness in both single-GPU and distributed environments. Our experiments, conducted on datasets of varying sizes and across multiple GPUs, broaden the applicability of our work to various GNN applications that handle input datasets with irregular graph sizes.

Lee, Claire Songhyun↗
NASA NTRS1975

Shuttle avionics system

The avionics system of the Space Shuttle is designed in a fail operational/fail safe architecture. The guidance, navigation and control system is implemented, through the onboard Orbiter digital computers. Guidance, navigation and control sensors are triplex, while the flight control effectors are mechanized either in load sharing or quad structure. Two sets of basic flight instruments and controls are provided along with electronic interfaces to allow for multiple selection of input destination and display source selection. Communications, tracking and instrumentation subsystems are mechanized as a dual hardware design for key operational elements. The data processing system allows for quad, triplex, dual or single computer operation. The power distribution subsystem provides a triple bus system with appropriate tie elements. A functional description is given of the computer system, the data bus, the mass memory unit, the multiplexer/demultiplexer and the CRT display system.

Gardiner, R. A.↗
NASA NTRS2010

Kepler Science Operations Center Pipeline Framework

The Kepler mission is designed to continuously monitor up to 170,000 stars at a 30 minute cadence for 3.5 years searching for Earth-size planets. The data are processed at the Science Operations Center (SOC) at NASA Ames Research Center. Because of the large volume of data and the memory and CPU-intensive nature of the analysis, significant computing hardware is required. We have developed generic pipeline framework software that is used to distribute and synchronize the processing across a cluster of CPUs and to manage the resulting products. The framework is written in Java and is therefore platform-independent, and scales from a single, standalone workstation (for development and research on small data sets) to a full cluster of homogeneous or heterogeneous hardware with minimal configuration changes. A plug-in architecture provides customized control of the unit of work without the need to modify the framework itself. Distributed transaction services provide for atomic storage of pipeline products for a unit of work across a relational database and the custom Kepler DB. Generic parameter management and data accountability services are provided to record the parameter values, software versions, and other meta-data used for each pipeline execution. A graphical console allows for the configuration, execution, and monitoring of pipelines. An alert and metrics subsystem is used to monitor the health and performance of the pipeline. The framework was developed for the Kepler project based on Kepler requirements, but the framework itself is generic and could be used for a variety of applications where these features are needed.

Klaus, Todd C.↗
DOE OSTI2021

An adaptive Hessian approximated stochastic gradient MCMC method

Bayesian approaches have been successfully integrated into training deep neural networks. One popular family is stochastic gradient Markov chain Monte Carlo methods (SG-MCMC), which have gained increasing interest due to their ability to handle large datasets and the potential to avoid overfitting. Although standard SG-MCMC methods have shown great performance in a variety of problems, they may be inefficient when the random variables in the target posterior densities have scale differences or are highly correlated. Here, we present an adaptive Hessian approximated stochastic gradient MCMC method to incorporate local geometric information while sampling from the posterior. The idea is to apply stochastic approximation (SA) to sequentially update a preconditioning matrix at each iteration. The preconditioner possesses second-order information and can guide the random walk of a sampler efficiently. Instead of computing and saving the full Hessian of the log posterior, we use limited memory of the samples and their stochastic gradients to approximate the inverse Hessian-vector multiplication in the updating formula. Moreover, by smoothly optimizing the preconditioning matrix via SA, our proposed algorithm can asymptotically converge to the target distribution with a controllable bias under mild conditions. To reduce the training and testing computational burden, we adopt a magnitude-based weight pruning method to enforce the sparsity of the network. Our method is user-friendly and demonstrates better learning results compared to standard SG-MCMC updating rules. The approximation of inverse Hessian alleviates storage and computational complexities for large dimensional models. Numerical experiments are performed on several problems, including sampling from 2D correlated distribution, synthetic regression problems, and learning the numerical solutions of heterogeneous elliptic PDE. The numerical results demonstrate great improvement in both the convergence rate and accuracy.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Efficient Distributed Sequence Parallelism for Transformer-Based Image Segmentation

We introduce an efficient distributed sequence parallel approach for training transformer-based deep learning image segmentation models. The neural network models are comprised of a combination of a Vision Transformer encoder with a convolutional decoder to provide image segmentation mappings. The utility of the distributed sequence parallel approach is especially useful in cases where the tokenized embedding representation of image data are too large to fit into standard computing hardware memory. To demonstrate the performance and characteristics of our models trained in sequence parallel fashion compared to standard models, we evaluate our approach using a 3D MRI brain tumor segmentation dataset. We show that training with a sequence parallel approach can match standard sequential model training in terms of convergence. Furthermore, we show that our sequence parallel approach has the capability to support training of models that would not be possible on standard computing resources.

Lyngaas, Isaac↗
NASA NTRS2012

Autonomous Information Unit for Fine-Grain Data Access Control and Information Protection in a Net-Centric System

As communication and networking technologies advance, networks will become highly complex and heterogeneous, interconnecting different network domains. There is a need to provide user authentication and data protection in order to further facilitate critical mission operations, especially in the tactical and mission-critical net-centric networking environment. The Autonomous Information Unit (AIU) technology was designed to provide the fine-grain data access and user control in a net-centric system-testing environment to meet these objectives. The AIU is a fundamental capability designed to enable fine-grain data access and user control in the cross-domain networking environments, where an AIU is composed of the mission data, metadata, and policy. An AIU provides a mechanism to establish trust among deployed AIUs based on recombining shared secrets, authentication and verify users with a username, X.509 certificate, enclave information, and classification level. AIU achieves data protection through (1) splitting data into multiple information pieces using the Shamir's secret sharing algorithm, (2) encrypting each individual information piece using military-grade AES-256 encryption, and (3) randomizing the position of the encrypted data based on the unbiased and memory efficient in-place Fisher-Yates shuffle method. Therefore, it becomes virtually impossible for attackers to compromise data since attackers need to obtain all distributed information as well as the encryption key and the random seeds to properly arrange the data. In addition, since policy can be associated with data in the AIU, different user access and data control strategies can be included. The AIU technology can greatly enhance information assurance and security management in the bandwidth-limited and ad hoc net-centric environments. In addition, AIU technology can be applicable to general complex network domains and applications where distributed user authentication and data protection are necessary. AIU achieves fine-grain data access and user control, reducing the security risk significantly, simplifying the complexity of various security operations, and providing the high information assurance across different network domains.

Chow, Edward T.↗
NASA NTRS2015

Pi-Sat: A Low Cost Small Satellite and Distributed Spacecraft Mission System Test Platform

Current technology and budget trends indicate a shift in satellite architectures from large, expensive single satellite missions, to small, low cost distributed spacecraft missions. At the center of this shift is the SmallSatCubesat architecture. The primary goal of the Pi-Sat project is to create a low cost, and easy to use Distributed Spacecraft Mission (DSM) test bed to facilitate the research and development of next-generation DSM technologies and concepts. This test bed also serves as a realistic software development platform for Small Satellite and Cubesat architectures. The Pi-Sat is based on the popular $35 Raspberry Pi single board computer featuring a 700Mhz ARM processor, 512MB of RAM, a flash memory card, and a wealth of IO options. The Raspberry Pi runs the Linux operating system and can easily run Code 582s Core Flight System flight software architecture. The low cost and high availability of the Raspberry Pi make it an ideal platform for a Distributed Spacecraft Mission and Cubesat software development. The Pi-Sat models currently include a Pi-Sat 1U Cube, a Pi-Sat Wireless Node, and a Pi-Sat Cubesat processor card.The Pi-Sat project takes advantage of many popular trends in the Maker community including low cost electronics, 3d printing, and rapid prototyping in order to provide a realistic platform for flight software testing, training, and technology development. The Pi-Sat has also provided fantastic hands on training opportunities for NASA summer interns and Pathways students.

Software↗
DOE OSTI2022

Technical note: Using long short-term memory models to fill data gaps in hydrological monitoring networks

Abstract. Quantifying the spatiotemporal dynamics in subsurface hydrological flows over a long time window usually employs a network of monitoring wells. However, such observations are often spatially sparse with potential temporal gaps due to poor quality or instrument failure. In this study, we explore the ability of recurrent neural networks to fill gaps in a spatially distributed time-series dataset. We use a well network that monitors the dynamic and heterogeneous hydrologic exchanges between the Columbia River and its adjacent groundwater aquifer at the U.S. Department of Energy's Hanford site. This 10-year-long dataset contains hourly temperature, specific conductance, and groundwater table elevation measurements from 42 wells with gaps of various lengths. We employ a long short-term memory (LSTM) model to capture the temporal variations in the observed system behaviors needed for gap filling. The performance of the LSTM-based gap-filling method was evaluated against a traditional autoregressive integrated moving average (ARIMA) method in terms of error statistics and accuracy in capturing the temporal patterns of river corridor wells with various dynamics signatures. Our study demonstrates that the ARIMA models yield better average error statistics, although they tend to have larger errors during time windows with abrupt changes or high-frequency (daily and subdaily) variations. The LSTM-based models excel in capturing both high-frequency and low-frequency (monthly and seasonal) dynamics. However, the inclusion of high-frequency fluctuations may also lead to overly dynamic predictions in time windows that lack such fluctuations. The LSTM can take advantage of the spatial information from neighboring wells to improve the gap-filling accuracy, especially for long gaps in system states that vary at subdaily scales. While LSTM models require substantial training data and have limited extrapolation power beyond the conditions represented in the training data, they afford great flexibility to account for the spatial correlations, temporal correlations, and nonlinearity in data without a priori assumptions. Thus, LSTMs provide effective alternatives to fill in data gaps in spatially distributed time-series observations characterized by multiple dominant frequencies of variability, which are essential for advancing our understanding of dynamic complex systems.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2010

Scaled CMOS Technology Reliability Users Guide

The desire to assess the reliability of emerging scaled microelectronics technologies through faster reliability trials and more accurate acceleration models is the precursor for further research and experimentation in this relevant field. The effect of semiconductor scaling on microelectronics product reliability is an important aspect to the high reliability application user. From the perspective of a customer or user, who in many cases must deal with very limited, if any, manufacturer's reliability data to assess the product for a highly-reliable application, product-level testing is critical in the characterization and reliability assessment of advanced nanometer semiconductor scaling effects on microelectronics reliability. A methodology on how to accomplish this and techniques for deriving the expected product-level reliability on commercial memory products are provided.Competing mechanism theory and the multiple failure mechanism model are applied to the experimental results of scaled SDRAM products. Accelerated stress testing at multiple conditions is applied at the product level of several scaled memory products to assess the performance degradation and product reliability. Acceleration models are derived for each case. For several scaled SDRAM products, retention time degradation is studied and two distinct soft error populations are observed with each technology generation: early breakdown, characterized by randomly distributed weak bits with Weibull slope (beta)=1, and a main population breakdown with an increasing failure rate. Retention time soft error rates are calculated and a multiple failure mechanism acceleration model with parameters is derived for each technology. Defect densities are calculated and reflect a decreasing trend in the percentage of random defective bits for each successive product generation. A normalized soft error failure rate of the memory data retention time in FIT/Gb and FIT/cm2 for several scaled SDRAM generations is presented revealing a power relationship. General models describing the soft error rates across scaled product generations are presented. The analysis methodology may be applied to other scaled microelectronic products and their key parameters.

Microelectronics Reliability↗