Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “HPC performance modeling”

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 271 records · Page 15

Position Papers for the ASCR Workshop on the Management and Storage of Scientific Data

The purpose of this workshop is to identify priority research directions in the area of data management for high-performance and scientific computing above and beyond HPC’s traditional "the parallel file system is the data-management system" model. Supporting the breadth of the DOE mission, including the explosion of AI uses and the growing needs of experimental and observational science, motivates revisiting our assumptions about data management. There are many facets of this topic to explore including: (1) Interfaces for accessing data that resides on traditional persistent storage as well as memory devices; (2) Storage-system architecture design that supports scientific workflows on varied hierarchical storage and networking devices; (3) Devising metadata management infrastructure to support FAIR principles (Findability, Accessibility, Interoperability, and Reusability); (4) Capturing provenance information about scientific data; (5) Utilizing AI to learn I/O patterns of emerging workloads for efficient data management; (6) Providing data management support for AI and complex workflows; and (7) Understanding the overlap between traditional storage systems and I/O (SSIO) efforts and data management. While the program committee has identified these topics as important areas for discussion, we welcome position papers from the community that propose additional topics of interest for discussion at the workshop. The workshop agenda will include breakout sessions for discussing these and selected topic areas to inform priority research directions for data management for high-performance and scientific computing.

97 MATHEMATICS AND COMPUTING↗

Report for the ASCR Workshop on the Management and Storage of Scientific Data

The purpose of this workshop is to identify priority research directions in the area of data management for high-performance and scientific computing above and beyond HPC’s traditional "the parallel file system is the data-management system" model. Supporting the breadth of the DOE mission, including the explosion of AI uses and the growing needs of experimental and observational science, motivates revisiting our assumptions about data management. There are many facets of this topic to explore including: (1) Interfaces for accessing data that resides on traditional persistent storage as well as memory devices; (2) Storage-system architecture design that supports scientific workflows on varied hierarchical storage and networking devices; (3) Devising metadata management infrastructure to support FAIR principles (Findability, Accessibility, Interoperability, and Reusability); (4) Capturing provenance information about scientific data; (5) Utilizing AI to learn I/O patterns of emerging workloads for efficient data management; (6) Providing data management support for AI and complex workflows; and (7) Understanding the overlap between traditional storage systems and I/O (SSIO) efforts and data management. While the program committee has identified these topics as important areas for discussion, we welcome position papers from the community that propose additional topics of interest for discussion at the workshop. The workshop agenda will include breakout sessions for discussing these and selected topic areas to inform priority research directions for data management for high-performance and scientific computing.

97 MATHEMATICS AND COMPUTING↗

From Data to Models and Analytics: An Integrated Scientific Computing Platform to Accelerate Environmental Research SBIR Phase I (Final Technical Report)

Phase I research was conducted to investigate the technical feasibility of an integrated open source platform GeoLib to provide turnkey solutions for model-data integration for environmental systems science (ESS), using modern web technologies, advances in HPC and cloud computing and high-performance visualization. A key feature of our platform is that it minimizes and automates data movement between data storage and computing systems. Another critical feature of the design is that it enables users to perform high-fidelity post-processing by performing high-performance visualization server-side and in-situ on the compute cluster close to the data. The investigation, which was performed over a 12-month period ending April 2019, centered on developing prototype software to demonstrate the key enabling technologies required for this system. This included: (i) rapid access to data from multiple remote sources, via a unified data interface; (ii) software for scalable geospatial visualization of raster, vector, and point cloud data on the web; (iii) development of an easy-to-use API for creating simulation and modeling workflows; (iv) remote data visualization and processing on middle-tier servers; and (v) remote data processing on high-performance computing systems. A critical part of this work is that all of these capabilities be accessible to end users through standard desktop tools such as Project Jupyter, which is widely adopted for ESS simulation and modeling.

97 MATHEMATICS AND COMPUTING↗

Performance Assessment of a Generic Nuclear Waste Repository in Shale - 18392

A performance assessment (PA) is an important component of a comprehensive safety analysis for a nuclear waste repository. In a PA, probabilistic simulations of the total repository system are performed, and results are evaluated against performance metrics. Uncertainty and sensitivity analyses help prioritize additional research and model development. The United States Department of Energy (DOE) has been developing a state of the art PA simulation software tool-kit, Geologic Disposal Safety Assessment (GDSA) Framework, that couples increasingly higher fidelity models of subsystem processes into total system PA simulations. Over the past several years, PAs of generic repositories in three geologic media (salt, shale, and crystalline rock) have demonstrated ongoing developments in capability. The current PA of a nuclear waste repository in a generic shale formation showcases GDSA Framework, including capabilities in domain discretization (Cubit), multi-physics simulations (PFLOTRAN), uncertainty and sensitivity analysis (Dakota), and visualization (Paraview). The generic shale reference case considers the disposal of 22,000 metric tons heavy metal of commercial spent nuclear fuel (SNF) in a generic shale formation. PA simulations account for the thermal load and radionuclide inventory of the waste form, components of the engineered barrier system including waste package and buffer, and components of the natural barrier system including the host rock shale and underlying and overlying stratigraphic units. Two repository layouts are considered, one for emplacement of waste packages containing 12 pressurized water reactor (PWR) assemblies, and one for 4-PWR waste package emplacement. Model domains are half-symmetry and contain between 7 and 22 million grid cells. Grid refinement captures the detail of individual waste packages, emplacement drifts, access drifts, and shafts. Simulations are run in a high-performance computing (HPC) environment on 512 or 2048 processes. The governing equations describing coupled heat and fluid flow and reactive transport are solved with PFLOTRAN, an open-source, massively parallel multiphase flow and reactive transport code. Additional simulated processes include waste package degradation; waste form dissolution; radioactive decay and ingrowth in aqueous, solid, and adsorbed phases; and a simple biosphere based on a pumping well. Simulations are run to 106 y; dose and radionuclide concentrations are observed within aquifers at a point approximately 5 km downgradient of the repository. Dakota is used to sample likely ranges of input parameters including waste form and waste package degradation rates and properties of engineered and natural materials to quantify uncertainty in predicted concentrations and sensitivity to input parameters. Given the assumptions of the reference case, results of 12-PWR and 4-PWR simulations are very similar. Dose at the pumping well 5-km downgradient of the repository does not exceed 10-9 Sv/y. I-129 is the largest contributor to dose. The dose due to Cl-36, the next largest contributor, is orders of magnitude smaller. At the pumping well, the concentration of I-129 5-km downgradient does not exceed 10-14 mol/L. Rank correlation coefficients indicate that the maximum concentration of I-129 within the aquifer is sensitive to shale (repository host rock) porosity and aquifer permeability. These results affirm that HPC-capable codes can be used to simulate important multi-physics couplings directly in a total system performance assessment of a geologic repository, and can be used in prioritization of future research and model development. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

A Faster-Than-Real-Time Framework for Reliability-Oriented Simulation of PV Inverters

Physics-of-Failure (PoF) based reliability assessment for photovoltaic (PV) inverters requires long-duration electrical and electrothermal stress histories, yet generating such stress histories with high-fidelity switching models over year long mission profiles is computationally prohibitive. Conventional methods either sacrifice modeling fidelity for speed or require runtimes that are impractical for design iteration and uncertainty studies. To address this bottleneck, this paper presents a High-Performance Computing (HPC) based simulation frame work for faster-than-real-time reliability-oriented simulation. The proposed framework integrates the Average-to-Switching (A2S) method with parallel computing techniques to accelerate switching-level waveform reconstruction. We further introduce optimization strategies, including cluster merging and sensitivity based mission profile screening, to reduce the computational burden. Evaluated using real-world mission profile inputs and a MATLAB/Simulink switching-model reference, the framework reduces the simulation time for a one-year mission from an intractable multi-year duration to approximately 7.3 minutes while maintaining low waveform error. This acceleration provides a practical reliability-oriented simulation engine that can be coupled with component-specific aging models for subsequent PV inverter PoF assessment.

High-performance Computing↗

Earthquake Ground Motion Simulations on the Sierra High-Performance Computing System

Physics-based three-dimensional (3D) numerical simulations of earthquake ground motions provide an alternative to reliance on limited and highly scattered empirical data and models for seismic hazard assessment. We performed fully 3D earthquake ground motion simulations on the Sierra and Lassen high-performance computing (HPC) platforms at Livermore Computing. These new HPC platforms feature graphics processing unit (GPU) acceleration for improved performance. Ground motions were computed with SW4, a summation-by-parts finite difference code for anelastic wave propagation. SW4 was recently enhanced to run efficiently on GPU-accelerated platforms. We found that SW4 runs almost fifty (50) times faster in terms of node-hours on new GPU-accelerated platforms compared to conventional central processing unit (CPU) platforms with agreement of computed ground motion time-series to the full single precision which they are written. We focused on moment magnitude, M W , 7.0 earthquakes on the Hayward Fault. We were able to run three simulations resolving ground motions from static displacements (0 Hz) to frequencies of 10 Hz (wavelengths as short as 50 meters) making these Sierra calculations the highest resolution regional-scale earthquake simulations ever made for California. Results show good agreement between simulated ground motion intensities (peak ground velocity and acceleration, spectral response) and empirical models, including frequencies above 10 Hz. On Lassen we were able to run a suite of 52 M W 7.0 Hayward Fault earthquake simulations resolved 0-5 Hz for both plane-layered, one-dimensional (1D) and 3D Earth models. These simulations were used to develop a methodology for characterizing and removing systematic path and site effects which bias ground motion intensities. These simulations using the newly enhanced SW4 code allow us to run higher resolution seismic simulations with shorter run times, providing a new capability for seismic hazard and risk studies. Results were presented at scientific conferences and workshops and two publications are being prepared based on work performed in this project.

58 GEOSCIENCES↗

HPC Annual Report 2017

Sandia National Laboratories has a long history of significant contributions to the high performance community and industry. Our innovative computer architectures allowed the United States to become the first to break the teraFLOP barrier—propelling us to the international spotlight. Our advanced simulation and modeling capabilities have been integral in high consequence US operations such as Operation Burnt Frost. Strong partnerships with industry leaders, such as Cray, Inc. and Goodyear, have enabled them to leverage our high performance computing (HPC) capabilities to gain a tremendous competitive edge in the marketplace. As part of our continuing commitment to providing modern computing infrastructure and systems in support of Sandia missions, we made a major investment in expanding Building 725 to serve as the new home of HPC systems at Sandia. Work is expected to be completed in 2018 and will result in a modern facility of approximately 15,000 square feet of computer center space. The facility will be ready to house the newest National Nuclear Security Administration/Advanced Simulation and Computing (NNSA/ASC) Prototype platform being acquired by Sandia, with delivery in late 2019 or early 2020. This new system will enable continuing advances by Sandia science and engineering staff in the areas of operating system R&D, operation cost effectiveness (power and innovative cooling technologies), user environment and application code performance.

97 MATHEMATICS AND COMPUTING↗

Electromagnetic Modeling of Lattice Structures in Additively Manufactured Electric Machines

Recent advancements in metal additive manufacturing (AM) show great potential to revolutionize the design and manufacturing of electromagnetic components used in the field of electrical engineering. Lattice structures directly printed by AM processes typically offer better structural performance with reduced weight, such as high stiffness, surface area, elongation, energy absorption, and porosity, than the solid counterpart. This paper aims to study the electromagnetic modeling of lattice structures used in additively manufactured magnetic cores or windings for electric machines. Three dimensional (3D) electromagnetic finite element (FE) analysis with high performance computing (HPC) shows the highest fidelity in predicting the electromagnetic performance of designs with lattice structures by preserving complex geometry details. FE-based homogenization methods have also been explored to potentially speed up concept design. A case study based on an additively manufactured axial-flux permanent magnet machine with a Hilbert pattern stator validates the discussed electromagnetic modeling approaches.

Chung, Shi-Uk [ANSYS, Inc.]↗

Data readiness pipeline patterns for scientific AI at scale: Insights from climate, fusion, life sciences, and materials

This article examines how data readiness for AI principles apply to large scientific datasets used to train foundation models. We analyze archetypal workflows across four representative domains—climate, nuclear fusion, life sciences, and materials—to identify common preprocessing patterns and domain‐specific constraints. We introduce a two‐dimensional readiness model that combines canonical preprocessing patterns with a five‐level operational readiness scale, both tailored to high‐performance computing (HPC) environments. This construct helps outline key challenges in transforming large‐scale scientific data into formats suitable for scalable AI training. Together, these dimensions form a conceptual maturity matrix that characterizes scientific data readiness and guides infrastructure development toward standardized, cross‐domain support for scalable and reproducible AI for science. Finally, we evaluate this maturity matrix in the context of case studies including ClimaX (climate), AFLOW (materials), OpenFold (proteomics), and DIII‐D fusion disruption‐prediction workflows, from which we distill lessons learned and provide recommendations to guide practitioners in developing robust AI‐readiness pipelines. Finally, we discuss remaining cross‐cutting challenges that persist across scientific domains.

97 MATHEMATICS AND COMPUTING↗

Scalable PGAS Metadata Management on Extreme Scale Systems

Programming models intended to run on exascale systems have a number of challenges to overcome, specially the sheer size of the system as measured by the number of concurrent software entities created and managed by the underlying runtime. It is clear from the size of these systems that any state maintained by the programming model has to be strictly sub-linear in size, in order not to overwhelm memory usage with pure overhead. A principal feature of Partitioned Global Address Space (PGAS) models is providing easy access to global-view distributed data structures. In order to provide efficient access to these distributed data structures, PGAS models must keep track of metadata such as where array sections are located with respect to processes/threads running on the HPC system. As PGAS models and applications become ubiquitous on very large transpetascale systems, a key component to their performance and scalability will be efficient and judicious use of memory for model overhead (metadata) compared to application data. We present an evaluation of several strategies to manage PGAS metadata that exhibit different space/time tradeoffs. We use two real-world PGAS applications to capture metadata usage patterns and gain insight into their communication behavior.

Global Arrays↗

A Study of Complex Deep Learning Networks on High Performance, Neuromorphic, and Quantum Computers

Current Deep Learning models use highly optimized convolutional neural networks (CNN) trained on large graphical processing units (GPU)-based computers with a fairly simple layered network topology, i.e., highly connected layers, without intra-layer connections. Complex topologies have been proposed, but are intractable to train on current systems. Building the topologies of the deep learning network requires hand tuning, and implementing the network in hardware is expensive in both cost and power.In this paper, we evaluate deep learning models using three different computing architectures to address these problems: quantum computing to train complex topologies, high performance computing (HPC) to automatically determine network topology, and neuromorphic computing for a low-power hardware implementation. Due to input size limitations of current quantum computers we use the MNIST dataset for our evaluation.The results show the possibility of using the three architectures in tandem to explore complex deep learning networks that are untrainable using a von Neumann architecture. We show that a quantum computer can find high quality values of intra-layer connections and weights, while yielding a tractable time result as the complexity of the network increases; a high performance computer can find optimal layer-based topologies; and a neuromorphic computer can represent the complex topology and weights derived from the other architectures in low power memristive hardware.This represents a new capability that is not feasible with current von Neumann architecture. It potentially enables the ability to solve very complicated problems unsolvable with current computing technologies.

97 MATHEMATICS AND COMPUTING↗

Flexible silicon photonic architecture for accelerating distributed deep learning

The increasing size and complexity of deep learning (DL) models have led to the wide adoption of distributed training methods in datacenters (DCs) and high-performance computing (HPC) systems. However, communication among distributed computing units (CUs) has emerged as a major bottleneck in the training process. In this study, we propose Flex-SiPAC, a flexible silicon photonic accelerated compute cluster designed to accelerate multi-tenant distributed DL training workloads. Flex-SiPAC takes a co-design approach that combines a silicon photonic hardware platform with a tailored collective algorithm, optimized to leverage the unique physical properties of the architecture. The hardware platform integrates a novel wavelength-reconfigurable transceiver design and a micro-resonator-based wavelength-reconfigurable switch, enabling the system to achieve flexible bandwidth steering in the wavelength domain. The collective algorithm is designed to support reconfigurable topologies, enabling efficient all-reduce communications that are commonly used in DL training. The feasibility of the Flex-SiPAC architecture is demonstrated through two testbed experiments. First, an optical testbed experiment demonstrates the flexible routing of wavelengths by shuffling an array of input wavelengths using a custom-designed spatial-wavelength selective switch. Second, a four-GPU testbed running two DL workloads shows a 23% improvement in job completion time compared to a similarly sized leaf-spine topology. We further evaluate Flex-SiPAC using large-scale simulations, which show that Flex-SiPAC is able to reduce the communication time by 26% to 29% compared to state-of-the-art compute clusters under representative collective operations.

Wu, Zhenguo (ORCID:0000000322847985)↗

A Study of Complex Deep Learning Networks on High Performance, Neuromorphic, and Quantum Computers

Current Deep Learning models use highly optimized convolutional neural networks (CNN) trained on large graphical processing units (GPU)-based computers with a fairly simple layered network topology, i.e., highly connected layers, without intra-layer connections. Complex topologies have been proposed, but are intractable to train on current systems. Building the topologies of the deep learning network requires hand tuning, and implementing the network in hardware is expensive in both cost and power. In this paper, we evaluate deep learning models using three different computing architectures to address these problems: quantum computing to train complex topologies, high performance computing (HPC) to automatically determine network topology, and neuromorphic computing for a low-power hardware implementation. Due to input size limitations of current quantum computers we use the MNIST dataset for our evaluation. The results show the possibility of using the three architectures in tandem to explore complex deep learning networks that are untrainable using a von Neumann architecture. We show that a quantum computer can find high quality values of intra-layer connections and weights, while yielding a tractable time result as the complexity of the network increases; a high performance computer can find optimal layer-based topologies; and a neuromorphic computer can represent the complex topology and weights derived from the other architectures in low power memristive hardware. This represents a new capability that is not feasible with current von Neumann architecture. It potentially enables the ability to solve very complicated problems unsolvable with current computing technologies.

Potok, Thomas E↗

REVEAL: An Extensible Reduced Order Model Builder for Simulation and Modeling

Many science domains need to build computationally efficient and accurate representations of high fidelity, computationally expensive simulations. These computationally efficient versions are known as reduced-order models. This paper presents the design and implementation of a novel reduced-order model (ROM) builder, the REVEAL toolset. This toolset generates ROMs based on science- and engineering-domain specific simulations executed on high performance computing (HPC) platforms. The toolset encompasses a range of sampling and regression methods that can be used to generate a ROM, automatically quantifies the ROM accuracy, and provides support for an iterative approach to improve ROM accuracy. REVEAL is designed to be extensible in order to utilize the core functionality with any simulator that has published input and output formats. It also defines programmatic interfaces to include new sampling and regression techniques so that users can ‘mix and match’ mathematical techniques to best suit the characteristics of their model. In this paper, we describe the architecture of REVEAL and demonstrate its usage with a computational fluid dynamics model used in carbon capture.

Surrogate model, reduced order model, sensitivity ↗

The Minos Computing Library: Efficient Parallel Programming for Extremely Heterogeneous Systems

Hardware specialization has become the silver bullet to achieve efficient high performance, from Systems-on-Chip systems, where hardware specialization can be ``extreme'', to large-scale HPC systems. As the complexity of the systems increases, so does the complexity of programming such architectures in a portable way. This work introduces the Minos Computing Library (MCL), as system software, programming model, and programming model runtime that facilitate programming extremely heterogeneous systems. MCL supports the execution of several multi-threaded applications within the same compute node, performs asynchronous execution of application tasks, efficiently balances computation across hardware resources, and provides performance portability. We show that code developed on a personal desktop automatically scales up to fully utilize powerful workstations with 8 GPUs and down to power-efficient embedded systems. MCL provides up to 17.5x speedup over OpenCL on NVIDIA DGX-1 systems and up to 1.88x speedup on single-GPU systems. In multi-application workloads, MCL dynamically resource allocation provides up to 2.43x performance improvement over manual, static allocation of computing resources.

Gioiosa, Roberto↗

Understanding the Impact of Dynamic Power Capping on Application Progress

Electrical power has become an important design constraint in high-performance computing (HPC) systems. On future HPC machines, power is likely to be a budgeted resource and thus managed dynamically. Power management software needs to reliably measure application performance at runtime in order to respond effectively to changes in application behavior. Execution time tells us little about how the science in the application is progressing toward an application-defined end goal. To the best of our knowledge, no study has defined or categorized online application progress in the context of power management. Based on semi-structured interviews with HPC application-specialists, we define an online notion of progress—an application-specific metric that can be monitored at runtime to provide a sense of the rate at which application science is being performed. Using instrumentation, we characterize and categorize the progress of various production scientific applications and benchmarks. We propose a model of the impact of dynamic power capping on application progress. By experimental evaluation, we show that our model accurately captures the general behavior of the progress of different classes of applications under a power cap. We believe that such a model is an important first step toward the design of more dynamic power management policies for HPC systems.

application progress↗

Advanced coal gasifier designs using large-scale simulations

Porting of the legacy code MFIX to a high performance computer (HPC) and the useof high resolution simulations for the design of a coal gasifier are described here. MFIX isbased on a continuum multiphase flow model that considers gas and solids to forminterpenetrating continua. Low resolution simulations of a commercial scale gasifier with avalidated MFIX model revealed interesting physical phenomena with implications on thegasifier design, which prompted the study reported here. To be predictive, the simulations needto model the spatiotemporal variations in gas and solids volume fractions, velocities,temperatures with any associated phase change and chemical reactions. These processes occurat various time- and length-scales requiring very high spatial resolution and large number ofiterations with small time-steps. We were able to perform perhaps the largest knownsimulations of gas-solids reacting flows, providing detailed information about the gas-solidsflow structure and the pressure, temperature and species distribution in the gasifier. One keyfinding is the new features of the coal jet trajectory revealed with the high spatial resolution,which provides information on the accuracy of the lower resolution simulations.Methodologies for effectively combining high and low resolution simulations for designstudies must be developed. From a computational science perspective, we found that globalcommunication has to be reduced to achieve scalability to 1000s of cores, hybridparallelization is required to effectively utilize the multicore chips, and the wait time in thebatch queue significantly increases the actual time-to-solution. From our experience,development is required in the following areas: efficient solvers for heterogeneous, massivelyparallel systems; data analysis tools to extract information from large data sets; andprogramming environments for easily porting legacy codes to HPC.

01 COAL, LIGNITE, AND PEAT↗

Scalable Federated Learning for Scientific Foundation Models on Leadership-Class Systems

Federated learning (FL) at leadership-class HPC systems remains largely unexplored, despite growing interest in deploying federated workflows on modern HPC systems. This paper provides the first system-level empirical characterization of federated fine-tuning of pretrained foundation models on an exascale supercomputer under a multi-node deployment. Using up to 96 concurrent FL clients deployed across Frontier nodes, we study the impact of client scale, model size, data heterogeneity, partial participation, and differential privacy on runtime, communication overhead, and convergence stability. Our results show that pretrained transformer models remain robust to heterogeneity, client dropout, and privacy noise, while system efficiency degrades rapidly with scale as synchronizat and orchestration dominate runtime. We further demonstrate that system-aware execution strategies, including intra-node aggregation and early aggregation, significantly reduce wall-clock time without degrading model quality. These findings establish a practical performance baseline and inform the design of communication-efficient FL systems on leadership-class HPC platforms.

Kotevska, Olivera [ORNL] (ORCID:0000000316772243)↗