SEARCH · Engineering Papers

Results for “Computer Systems”

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.

SourceResearch connectionFrom yearThrough year

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

At least 847 records · Page 47

All sources NASA NTRS DOE OSTI

DOE OSTI2013

Software design space exploration for exascale combustion co-design

The design of hardware for next-generation exascale computing systems will require a deep understanding of how software optimizations impact hardware design trade-offs. In order to characterize how co-tuning hardware and software parameters affects the performance of combustion simulation codes, we created ExaSAT, a compiler-driven static analysis and performance modeling framework. Our framework can evaluate hundreds of hardware/software configurations in seconds, providing an essential speed advantage over simulators and dynamic analysis techniques during the co-design process. Our analytic performance model shows that advanced code transformations, such as cache blocking and loop fusion, can have a significant impact on choices for cache and memory architecture. Our modeling helped us identify tuned configurations that achieve a 90% reduction in memory traffic, which could significantly improve performance and reduce energy consumption. These techniques will also be useful for the development of advanced programming models and runtimes, which must reason about these optimizations to deliver better performance and energy efficiency.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2026

Augmenting LLM-Based Agents for Improved Performance in Pentesting and Commissioning Operational Technology in Critical Infrastructure

Artificial intelligence (AI), and more specifically large language models (LLMs) have the potential for use in penetration testing (“pentesting”) against devices, networks, and computer systems in information technology (IT). We explore the possibility of extending pentesting from IT systems to operational technology (OT) systems, which are more obscure than IT systems in their protocols and design. A challenge therefore exists when applying pretrained LLMs to OT systems as corpora are likely to underrepresent OT systems in comparison to other more prevalent systems. We evaluate augmentations of LLMs with various methods, especially retrieval augmented generation (RAG), to improve performance of the LLMs in the OT domain. In addition to pentesting, some of the testing of these OT devices may include commissioning to ensure that the newly installed devices work correctly. Our framework may also be applied in such cases.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Is Knowledge about Running Applications Helping Improve Runtime Prediction of HPC Jobs?

High-performance computing systems rely upon scheduling algorithms to achieve high utilization. These schedulers rely upon user estimates of job resource requirements, such as runtime, to determine optimal scheduling of incoming jobs. These user estimates, however, are prone to error. To mitigate this error, significant research has been directed at providing better estimates of job runtime, usually employing machine learning techniques. These techniques are dependent upon the input features selected. Among the possible features is the primary application used by the job. In a survey of more than 20 papers directed at improving runtime prediction, only four included primary application as an input feature. We focus this investigation specifically on the value of adding primary application as an input feature, and find that it does improve model performance, especially for jobs with longer runtimes, though this improvement varies based on the application used. We recommend further research to determine the cause of this variability as well as an optimal strategy for employing a mixture of models both including and not including primary application as a feature.

MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Establishing Malware Attribution and Binary Provenance Using Multicompilation Techniques

Malware is a serious problem for computer systems and costs businesses and customers billions of dollars a year in addition to compromising their private information. Detecting malware is particularly difficult because malware source code can be compiled in many different ways and generate many different digital signatures, which causes problems for most anti-malware programs that rely on static signature detection. Our project uses a convolutional neural network to identify malware programs but these require large amounts of data to be effective. Towards that end, we gather thousands of source code files from publicly available programming contest sites and compile them with several different compilers and flags. Building upon current research, we then transform these binary files into image representations and use them to train a long-term recurrent convolutional neural network that will eventually be used to identify how a malware binary was compiled. This information will include the compiler, version of the compiler and the options used in compilation, information which can be critical in determining where a malware program came from and even who authored it.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Is Knowledge About Running Applications Helping Improve Runtime Prediction of HPC Jobs?

feature selection↗

DOE OSTI2023

TF9 Dataset Analysis

Incident Overview: In the time between November 2, 2019 and November 11, 2019, WheelByte was plagued by breaches in security. These insecurities led to breaches in customer data, company data, and even the death of an employee, Matthew Swift. They have launched an investigation into the company’s computer systems in hopes to find the root cause. We have been provided with the following artifacts from WheelByte: memory images, disk images, network packet captures, and emails. We have found multiple cyber-system attacks against WheelByte. Our investigation lasted from July 13th - August 3rd, 2023. WheelByte allowed us to look at any and every file, and there were no restrictions on what we could or could not use in our investigation. By the end of our investigation, we have been able to deduce who is behind the attack, what they have done, and why they did it. A company that is closely related to WheelByte is called Slyde. Slyde sells electric scooters and it is known that the Chief Executive Officer (CEO) of Slyde, Kimberly Holmes, sees WheelByte as a threat to business, as Wheelbyte sells electric skateboards. We have been able to deduce that Slyde is likely behind many of the malicious attacks. We have seen exfiltration addresses to Slyde domains, along with other Slyde information within their malware. We can see lots of traffic to and from Slyde Internet Protocol (IP) addresses. This may be an attempt to cripple WheelByte’s productivity to remove Slyde’s competitor from the market.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2015

Dry coupled magnetostrictive transducers for robotic inspection of dry storage casks - 15304

The lack of a national repository for used nuclear fuel has lead to the usage of dry storage casks as the de facto long term storage solution, the usage of which is prevalent throughout the United States. Furthermore, as shown in figure 3, the usage of dry storage is only expected to increase over time. Because these casks were originally intended as only an interim measure, there is an urgent need to develop methods to monitor their structural health over the long-term. Additionally, when concerns about the susceptibility of these casks to cracking is considered, the need for inspection becomes even more apparent. The use of robotic devices to inspect the interior of these casks for signs of damage has been proposed, but in order for this to be realized, advances in sensing technology must occur. We develop a novel type of transducer for non-destructive evaluation to aid in these efforts. We construct a prototype sensor, demonstrate simple defect detection with it, and present a design for an inspection robot equipping these transducers. Ultrasonic inspection is a well known method of non-destructive evaluation which works by measuring the response of ultrasonic waves propagating through a material. In through-transmission inspection, a transmitter and receiver are used to generate and detect waves in the material to be inspected. The presence of a crack or defect in the material will scatter some of the transmitted wave's energy, reducing the strength of the response at the receiver. Magnetostrictive transducers are often used to generate and detect guided waves. These transducers depend on the magnetostrictive effect, where certain materials will deform in the presence of a magnetic field. The typical construction of one of these transducers is shown in figure 5, consisting of a magnetostrictive foil, meandering coil, and a biasing magnet. Materials typically used for the foil include Fe-Co and Terfenol-D. For wave generation, a high-frequency alternating current is supplied to the coil of the transducer, which generates an oscillating magnetic field that causes the foil to deform. For reception, the inverse occurs: deformation of the foil results in a current in the coil that can be detected. The deformation of the foil is usually coupled to the material to be inspected with an adhesive, and waves are transmitted through this bond. Obviously, an inspection robot cannot have its sensors bound to the surface of the cask. A robot is being designed for inspection, and it must deal with many challenges inside the cask, including tight confines, right-angle transitions, high temperatures, and radiation. For our initial design, we consider only the geometrical constraints, and present a design for a prototype that can be assembled from COTS components. The robot needs to be articulated to navigate the bends in the cask, so each wheel is individually actuated. The robot is controlled via a tether, which also supplies electrical power and compressed air. For localization, ultrasonic range-finders and a camera are equipped. Using the basic transducer, we experimented with various configurations of materials and geometry. In order to gather relevant data, we used force sensitive resistors to measure applied force, and a Matec computer system to generate and measure ultrasonic waves. A testing rig was constructed to simulate the geometrical constraints of the interior of the cask, and we experimented using deadweights, a pneumatic cylinder, and an air bladder to apply the force. It was discovered early in the testing process that the performance of the transducer was very sensitive to the alignment of the components, and the method by which the foil was attached. To maximize performance, we designed an enclosure that would hold the components in alignment while still allowing the foil to float. To test the effectiveness of our new transducer, we cut circular holes of varying diameter in 2 mm aluminum plate, and measured the response across them. It can be seen that there is a clear relationship between the strength of the received signal and the force applied, and to a lesser extent, the size of the defect. This design performs adequately as a proof of concept, and a more precisely fabricated enclosure should further increase the accuracy of its measurements.

42 ENGINEERING↗

DOE OSTI2024

Analysis and Benchmarking of feature reduction for classification under computational constraints

Abstract Machine learning is most often expensive in terms of computational and memory costs due to training with large volumes of data. Current computational limitations of many computing systems motivate us to investigate practical approaches, such as feature selection and reduction, to reduce the time and memory costs while not sacrificing the accuracy of classification algorithms. In this work, we carefully review, analyze, and identify the feature reduction methods that have low costs/overheads in terms of time and memory. Then, we evaluate the identified reduction methods in terms of their impact on the accuracy, precision, time, and memory costs of traditional classification algorithms. Specifically, we focus on the least resource intensive feature reduction methods that are available in Scikit-Learn library. Since our goal is to identify the best performing low-cost reduction methods, we do not consider complex expensive reduction algorithms in this study. In our evaluation, we find that at quadratic-scale feature reduction, the classification algorithms achieve the best trade-off among competitive performance metrics. Results show that the overall training times are reduced 61%, the model sizes are reduced 6×, and accuracy scores increase 25% compared to the baselines on average with quadratic scale reduction.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Software and computing for Run 3 of the ATLAS experiment at the LHC

The ATLAS experiment has developed extensive software and distributed computing systems for Run 3 of the LHC. These systems are described in detail, including software infrastructure and workflows, distributed data and workload management, database infrastructure, and validation. The use of these systems to prepare the data for physics analysis and assess its quality are described, along with the software tools used for data analysis itself. An outlook for the development of these projects towards Run 4 is also provided.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2021

Pre-conditioned BFGS-based uncertainty quantification in elastic full-waveform inversion

SUMMARY Full-waveform inversion has become an essential technique for mapping geophysical subsurface structures. However, proper uncertainty quantification is often lacking in current applications. In theory, uncertainty quantification is related to the inverse Hessian (or the posterior covariance matrix). Even for common geophysical inverse problems its calculation is beyond the computational and storage capacities of the largest high-performance computing systems. In this study, we amend the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm to perform uncertainty quantification for large-scale applications. For seismic inverse problems, the limited-memory BFGS (L-BFGS) method prevails as the most efficient quasi-Newton method. We aim to augment it further to obtain an approximate inverse Hessian for uncertainty quantification in FWI. To facilitate retrieval of the inverse Hessian, we combine BFGS (essentially a full-history L-BFGS) with randomized singular value decomposition to determine a low-rank approximation of the inverse Hessian. Setting the rank number equal to the number of iterations makes this solution efficient and memory-affordable even for large-scale problems. Furthermore, based on the Gauss–Newton method, we formulate different initial, diagonal Hessian matrices as pre-conditioners for the inverse scheme and compare their performances in elastic FWI applications. We highlight our approach with the elastic Marmousi benchmark model, demonstrating the applicability of pre-conditioned BFGS for large-scale FWI and uncertainty quantification.

58 GEOSCIENCES↗

DOE OSTI2022

Performance and power modeling and prediction using MuMMI and 10 machine learning methods

Energy-efficient scientific applications require insight into how high performance computing system features impact the applications' power and performance. This insight can result from the development of performance and power models. Here, in this article, we use the modeling and prediction tool MuMMI (Multiple Metrics Modeling Infrastructure) and 10 machine learning methods to model and predict performance and power consumption and compare their prediction error rates. We use an algorithm-based fault-tolerant linear algebra code and a multilevel checkpointing fault-tolerant heat distribution code to conduct our modeling and prediction study on the Cray XC40 Theta and IBM BG/Q Mira at Argonne National Laboratory and the Intel Haswell cluster Shepard at Sandia National Laboratories. Our experimental results show that the prediction error rates in performance and power using MuMMI are less than 10% for most cases. By utilizing the models for runtime, node power, CPU power, and memory power, we identify the most significant performance counters for potential application optimizations, and we predict theoretical outcomes of the optimizations. Based on two collected datasets, we analyze and compare the prediction accuracy in performance and power consumption using MuMMI and 10 machine learning methods.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2019

Statistical and machine learning models for optimizing energy in parallel applications

Rising power costs and constraints are driving a growing focus on the energy efficiency of high performance computing systems. The unique characteristics of a particular system and workload and their effect on performance and energy efficiency are typically difficult for application users to assess and to control. Settings for optimum performance and energy efficiency can also diverge, so we need to identify trade-off options that guide a suitable balance between energy use and performance. We present statistical and machine learning models that only require a small number of runs to make accurate Pareto-optimal trade-off predictions using parameters that users can control. We study model training and validation using several parallel kernels and more complex workloads, including Algebraic Multigrid (AMG), Large-scale Atomic Molecular Massively Parallel Simulator, and Livermore Unstructured Lagrangian Explicit Shock Hydrodynamics. Here, we demonstrate that we can train the models using as few as 12 runs, with prediction error of less than 10%. Our AMG results identify trade-off options that provide up to 45% improvement in energy efficiency for around 10% performance loss. We reduce the sample measurement time required for AMG by 90%, from 13 h to 74 min.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Scalable Data Management, Analysis, and Visualization (SDAV) Institute (Final Report)

The purpose of the SDAV institute is to provide tools and expertise in scientific data management, analysis, and visualization to application scientists on leadership-class computing facilities. Our goal is to actively work with application teams to assist them in achieving breakthrough science, and to provide technical solutions in the data management, analysis, and visualization regimes that are broadly used by the computational science community. Over the last 5 years members of our institute worked directly with application scientists to assist them by applying the best tools and technologies at our disposal. We also enhanced our tools based on input from scientists on their needs. Many of the applications we have been working with are based on connections with scientists established in previous years. However, we contacted additional scientists though our outreach activities, as well as engaging application teams running on leading DOE computing systems. Our approach is to employ an evolutionary development and deployment process: first considering the application of existing tools, followed by the customization necessary for each particular application, and then the deployment in real frameworks and infrastructures. The institute is organized into three areas, each with area leaders, who keep track of progress, engagement of application scientists, and results. The areas are: (1) Data Management, (2) Data Analysis, and (3) Visualization. This report is organized along these areas. However, often there are multiple technologies from these areas that are applied to a single application need. These are described on a case-by-case basis in the appropriate sections. This final report for the SDAV Institute covers the 5-year period from February 2012 to February 2017. This report covers activities in all SDAV institutions listed next. Laboratories: Argonne National Laboratory (ANL), Lawrence Berkeley National Laboratory (LBNL), Lawrence Livermore National Laboratory (LLNL), Los Alamos National Laboratory (LLNL), Oak Ridge National Laboratory (ORNL), Sandia National Laboratory (SNL). Universities: Georgia Institute of Technology, North Carolina State University, Northwestern University, Ohio State University, Oregon State University, Rutgers University, University of California at Davis, University of Utah. Industry: Kitware.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

QoS-aware edge AI placement and scheduling with multiple implementations in FaaS-based edge computing

Resource constraints on the computing continuum require that we make smart decisions for serving AI-based services at the network edge. AI-based services typically have multiple implementations (e.g., image classification implementations include SqueezeNet, DenseNet, and others) with varying trade-offs (e.g., latency and accuracy). The question then is how should AI-based services be placed across Function-as-a-Service (FaaS) based edge computing systems in order to maximize total Quality-of-Service (QoS). To address this question, we propose a problem that jointly aims to solve (i) edge AI service placement and (ii) request scheduling. These are done across two time-scales (one for placement and one for scheduling). Here we first cast the problem as an integer linear program. We then decompose the problem into separate placement and scheduling subproblems and prove that both are NP-hard. We then propose a novel placement algorithm that places services while considering device-to-device communication across edge clouds to offload requests to one another. Our results show that the proposed placement algorithm is able to outperform a state-of-the-art placement algorithm for AI-based services, and other baseline heuristics, with regard to maximizing total QoS. Additionally, we present a federated learning-based framework, FLIES, to predict the future incoming service requests and their QoS requirements. Our results also show that our FLIES algorithm is able to outperform a standard decentralized learning baseline for predicting incoming requests and show comparable predictive performance when compared to centralized training.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

High Fidelity CFD Simulations Supporting the KP-FHR

Kairos Power, LLC, is developing its version of the Fluoride-cooled High-temperature Reactor, the KP-FHR. The design uses a pebble bed core with fluoride salt as a coolant. The pebbles used in the KP-FHR have a diameter of 4 cm, with a shell fuel region where TRISO particles are embedded. A Pebble bed core design is adopted by several Gen IV reactors, They boast many benefits, such as fuel integrity, highly efficient heat transfer, and passive safety. However, it is challenging to accurately predict temperature and flow inside a pebble bed. Traditional approaches use the porous media model, which regards the pebble bed as a continuous medium, but with different temperature fields representing different levels, such as the fluid temperature, pebble surface temperature, and pebble center temperature. Empirical heat transfer correlations are adopted to calculate the heat transfer coefficient between different phases. However, empirical correlations are usually validated with experimental data, which usually lacks detail inside the pebble bed. The available experimental data is also generally at a high Reynolds number, which falls outside of the conditions of KP-FHR. Explicit computational fluid dynamics (CFD) simulations of randomly packed pebble beds have only become feasible recently. This is thanks to the rapid development of computational power and scalable algorithms. In this work, we used the Spectral Element Method (SEM) CFD code NekRS to simulate the randomly packed pebble bed in a cylindrical container. NekRS, which is the GPU variant of Nek5000, but refactored to utilize the computational power of GPUs using the OCCA library to run on hybrid architecture high performance computing systems. It was initially developed with the libParamunal library, but truncated and tuned for large-scale turbulence simulation. As a result, the SEM reaches higher precision with the same degrees of freedom by using a high-order Lagrange polynomial basis distributed on Gauss-Lobatto-Legendre quadrature inside each element, compared to lower-order methods, such the Finite Volume Method and Finite Element Method. The report is divided into five parts. We start with a general discussion of the pebble bed reactor, along with a specific investigation into the KP-FHR. The second part presents the numerical methodology. In the third part, we study a modular pebble bed with 1741 pebbles in a container of 7 pebble-diameter radius. Beyond LES simulations done by NekRS, we also leveraged the thermal radiation model in OpenFOAM to study heat transfer under no-forced-flow scenarios. Then, in the fourth part we simulated a pebble bed similar to the size of the Hermes Test Reactor. The total number of pebbles is in these simulations is 34,374. The container radius is 14 pebble-diameters. Finally, the report concludes in part five, with a discussion of future work.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2024

Performance of explicit and IMEX MRI multirate methods on complex reactive flow problems within modern parallel adaptive structured grid frameworks

Large-scale multiphysics simulations are computationally challenging due to the coupling of multiple processes with widely disparate time scales. The advent of exascale computing systems exacerbates these challenges since these systems enable ever-increasing size and complexity. In recent years, there has been renewed interest in developing multirate methods as a means to handle the large range of time scales, as these methods may afford greater accuracy and efficiency than more traditional approaches of using implicit-explicit (IMEX) and low-order operator splitting schemes. However, to date there have been few performance studies that compare different classes of multirate integrators on complex application problems. In this work, we study the performance of several newly developed multirate infinitesimal (MRI) methods, implemented in the SUNDIALS solver package, on two reacting flow model problems built on structured mesh frameworks. The first model revisits prior work on a compressible reacting flow problem with complex chemistry that is implemented using BoxLib but where we now include comparisons between a new explicit MRI scheme with the multirate spectral deferred correction (SDC) methods in the original paper. The second problem uses the same complex chemistry as the first problem, combined with a simplified flow model, but runs at a large spatial scale where explicit methods become infeasible due to stability constraints. Two recently developed IMEX MRI multirate methods are tested. These methods rely on advanced features of the AMReX framework on which the model is built, such as multilevel grids and multilevel preconditioners. The results from these two problems show that MRI multirate methods can offer significant performance benefits on complex multiphysics application problems and that these methods may be combined with advanced spatial discretization to compound the advantages of both.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Side-channel Leakage Assessment Metrics: A Case Study of GIFT Block Ciphers

Determination of an adequate level of security and providing subsequent mechanisms to achieve it, is one of the most pressing problems regarding embedded computing devices. While there are some solutions available for resource-rich computer systems, direct application of these solutions to resource-constrained environments are often unfeasible. The fundamental problem for such resource-constrained systems is the fact that current cryptographic algorithms utilize significant energy consumption and storage overhead. Both the cryptographic algorithm and its physical implementation affect the resilience of a cryptosystem against side-channel attacks. A side-channel attack represents a process that exploits leakages in order to extract sensitive information such as the key. This paper focuses on Correlation Power Analysis (CPA) which is side-channel attack based on the power consumption leakage. In 2016 the U.S. Commerce Department’s National Institute of Standards and Technology (NIST) initiated the call for proposals of new cryptographic algorithms to strengthen the cryptographic defense of networked devices against cyberattacks and to protect the data created by those innumerable device. This work evaluates S-boxes used by NIST candidates PICCOLO, GIFT, and PRESENT, as well as several S-box variants that demonstrated sufficient weaknesses against classical cryptanalysis, for a quantitative comparison in terms of resiliency to CPA attack. Three well-known theoretical metrics are evaluated: transparency order (TO and RTO), nonlinearity, and signal-to-noise (SNR) ratio, aiming to characterize the resistance of these S-boxes against adversaries exploiting physical leakages. Experimental results from attacks on an 8- bit XMEGA were obtained via the ChipWhisperer platform and of all the S-boxes evaluated, GIFT64 with a PICCOLO S-box was found to be the most susceptible to CPA. Results showed that variations in TO and RTO were not sufficient to ensure practical CPA resistance and that among S-boxes with equal non-linearity there were no significant differences in the TO and SNR variants.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Side-channel Leakage Assessment Metrics: A Case Study of GIFT Block Ciphers

97 MATHEMATICS AND COMPUTING↗