Search NASA⌕ Search

SEARCH · Search NASA

Results for “runtime 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.

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

At least 271 records · Page 15

Towards Native Execution of Deep Learning on a Leadership-Class HPC System

Large parallel machines generally offer the best parallel performance with "native execution" that is achieved using codes developed with the optimized compilers, communication libraries, and runtimes offered on the machines. In this paper, we report and analyze performance results from native execution of deep learning on a leadership-class high-performance computing (HPC) system. Using our new code called DeepEx, we present a study of the parallel speed up and convergence rates of learning achieved with native parallel execution. In the trade-off between computational parallelism and synchronized convergence, we first focus on maximizing parallelism while still obtaining convergence. Scaling results are reported from execution on up to 15,000 GPUs using two scientific data sets from atom microscopy and protein folding applications, and also using the popular ImageNet data set. In terms of the traditional measure of parallel speed up, excellent scaling is observed up to 12,000 GPUs. Additionally, accounting for convergence rates of deep learning accuracy or error, a deep learning-specific metric called "learning speed up" is also tracked. The performance results indicate the need to evaluate parallel deep learning execution in terms of learning speed up, and point to additional directions for improved exploitation of high-end HPC systems.

Yoginath, Srikanth↗

PaRSEC: A Software Framework for Performance and Productivity on Hybrid, Manycore Platforms

As the era of computer architectures dominated by serial processors ends, the convergence of several unprecedented challenges suggests that closing the longstanding "application–architecture performance gap" will become more challenging than ever. To address this problem, the Parallel Runtime Scheduling and Execution Control (PaRSEC) project created a modular software framework that achieved two major objectives: first, it built a task-based runtime capable of delivering portable performance to a wide range of science and engineering applications at all levels of the platform pyramid, including the upcoming 100 Pflop/s systems and then exascale; and second, it supported and facilitated the work of developers in migrating their legacy codes and writing entirely new ones for the emerging hybrid and massively parallel manycore processor system designs. PaRSEC will support multiple domain-specific languages capable of increasing the developers' productivity while also providing the runtime with the constructs and flexibility necessary to exploit the maximal parallelism from parallel applications. Extensive preliminary research in dense linear algebra showed convincingly that a parameterized task graph representation that symbolically describes the algorithm content can achieve the project's twofold objective within that domain. The research also strongly suggested that this powerful method could be generalized to a far-wider variety of applications.

97 MATHEMATICS AND COMPUTING↗

Using containers to speed up development, to run integration tests and to teach about distributed systems

GlideinWMS is a workload manager provisioning resources for many experiments including CMS and DUNE. The software is distributed both as native packages and specialized production containers. Following an approach used in other communities like web development we built our workspaces, system-like containers to ease development and testing. Developers can change the source tree or check out a different branch and quickly reconfigure the services to see the effect of their changes. In this paper, we’ll talk about what differentiates workspaces from other containers. We’ll describe our base system composed of three containers. A one-node cluster including a compute element and a batch system. A GlideinWMS Factory controlling pilot jobs. And a scheduler and Frontend, to submit jobs and provision resources. Additional containers can be used for optional components. This system can easily run on a laptop and we’ll share our evaluation of different container runtimes, with an eye for ease of use and performance. Finally, we’ll talk about our experience as developers and with students. The GlideinWMS workspaces are easily integrated with IDEs like VS Code, simplifying debugging and allowing development and testing of the system also when offline. They simplified the training and onboarding of new team members and Summer interns. And they were useful in workshops where students could have first-hand experience with the mechanisms and components that, in production, run millions of jobs.

Mambelli, Marco↗

Using Containers to Speed Up Development, to Run Integration Tests and to Teach About Distributed Systems

GlideinWMS is a workload manager provisioning resources for many experiments, including CMS and DUNE. The software is distributed both as native packages and specialized production containers. Following an approach used in other communities like web development, we built our workspaces, system-like containers to ease development and testing. Developers can change the source tree or check out a different branch and quickly reconfigure the services to see the effect of their changes. In this paper, we will talk about what differentiates workspaces from other containers. We will describe our base system, composed of three containers: a one-node cluster including a compute element and a batch system, a GlideinWMS Factory controlling pilot jobs, and a scheduler and Frontend to submit jobs and provision resources. Additional containers can be used for optional components. This system can easily run on a laptop, and we will share our evaluation of different container runtimes, with an eye for ease of use and performance. Finally, we will talk about our experience as developers and with students. The GlideinWMS workspaces are easily integrated with IDEs like VS Code, simplifying debugging and allowing development and testing of the system even when offline. They simplified the training and onboarding of new team members and summer interns. And they were useful in workshops where students could have first-hand experience with the mechanisms and components that, in production, run millions of jobs.

Mambelli, Marco [Fermilab] (ORCID:0000000294892681↗

Development of near-optimal advanced control sequences for chiller plants with water-side economizers in U.S. Climates (ASHRAE RP-1661)

Various advanced control sequences for chiller plants with water-side economizers (WSE) have been proposed in literature, but the evaluation and optimization of those controls is limited. It is possible to maximize energy savings by selecting different sequences and related parameters based on the plant configuration, load, and climate. This paper addresses this gap by developing near-optimal advanced control sequences for chiller plants with WSEs. First, advanced control sequences for chiller plants with WSEs are categorized into condenser water, chilled water, and hybrid controls and representative sequences from each category are identified. Next, 504 different scenarios are optimized. These scenarios represent all possible combinations of two plant configurations, a constant or variable load profile, three advanced control sequences, and seven optimization parameter combinations in six climate zones. The results show the recommended near-optimal sequences can reduce energy consumption by up to 15% relative to the baseline depending on the configuration, load profile, and climate. Specifically, the CW-CHW sequence is recommended for the majority of systems because it is often the most energy efficient and/or reduces the runtime of chillers. The methodology in this paper provides practical guidance for achieving energy savings through near-optimal control of chiller plants with WSEs.

42 ENGINEERING↗

Extending Skel to Support the Development and Optimization of Next Generation I/O Systems

As the memory and storage hierarchy get deeper and more complex, it is important to have new benchmarks and evaluation tools that allow us to explore the emerging middleware solutions to use this hierarchy. Skel is a tool aimed at automating and refining this process of studying HPC I/O performance. It works by generating application I/O kernel/benchmarks as determined by a domain-specific model. This paper provides some techniques for extending Skel to address new situations and to answer new research questions. For example, we document use cases as diverse as using Skel to troubleshoot I/O performance issues for remote users, refining an I/O system model, and facilitating the development and testing of a mechanism for runtime monitoring and performance analytics. We also discuss data oriented extensions to Skel to support the study of compression techniques for Exascale scientific data management.

97 MATHEMATICS AND COMPUTING↗

Adapting Quantum Approximation Optimization Algorithm (QAOA) for Unit Commitment

In the present Noisy Intermediate-Scale Quantum (NISQ), hybrid algorithms that leverage classical resources to reduce quantum costs are particularly appealing. We formulate and apply such a hybrid quantum-classical algorithm to a power system optimization problem called Unit Commitment, which aims to satisfy a target power load at minimal cost. Our algorithm extends the Quantum Approximation Optimization Algorithm (QAOA) with a classical minimizer in order to support mixed binary optimization. Using Qiskit, we simulate results for sample systems to validate the effectiveness of our approach. Here, we also compare to purely classical methods. Our results indicate that classical solvers are effective for our simulated Unit Commitment instances with fewer than 400 power generation units. However, for larger problem instances, the classical solvers either scale exponentially in runtime or must resort to coarse approximations. This opens the door to potential quantum advantage for systems with several hundred units, though quantum error correction may be necessary at this scale.

42 ENGINEERING↗

Optimal operation of multi-plant steam district heating systems for enhanced efficiency and sustainability

Despite their crucial role in supplying heat and power to universities, industries, and healthcare facilities, many steam-based district heating systems rely on outdated control methods. Among these, multi-central plant districts are particularly challenging due to the complexities of coordinating multiple plants, optimizing load distributions, and managing system downtime. In response, new operational strategies are developed to enhance the efficiency and sustainability of steam districts while utilizing existing resources. These strategies include reducing plant operational pressure without compromising the reliable supply to buildings and optimizing load allocation across multiple plants. The load allocation considers boiler part-load efficiency, runtime, network losses, and building pressure set points, and is compared with traditional multi-boiler controls. To support this exploration, new dynamic Modelica models are developed. In addition, methods to reduce modeling complexities are incorporated, enhancing their suitability for practical applications. A holistic district-wide analysis using a real university case study demonstrates a 4.7% fuel savings by lowering boiler operational pressure from 900 kPa to 600 kPa, along with a 13.3% reduction in condensation losses across the distribution network. Furthermore, the load allocation approach results in a 13.1% reduction in fuel consumption during peak winter periods and 15.3% during shoulder periods, with corresponding decreases in carbon emissions and fuel costs. This approach can also save maintenance costs by reducing the boiler runtime by 49.6%. In conclusion, this research underscores the benefits of retrofitting aging steam district heating systems, offering immediate operational improvements by enhancing efficiency, meeting regulatory compliance, and extending infrastructure lifespans while delaying costly overhauls.

24 POWER TRANSMISSION AND DISTRIBUTION↗

PIPER: Performance Insight for Programmers and Exascale Runtimes: Guiding the Development of the Exascale Software Stack

The PIPER project set out to develop methodologies and software for measurement, analysis, attribution, and presentation of performance data for extreme-scale systems. Goals of the project were to support analysis of massive multi-scale parallelism, heterogeneous architectures, multi-faceted performance concerns, and to support both post-mortem performance analysis to identify program features that contribute to problematic performance and on-line performance analysis to drive adaptation. This final report summarizes the research and development activity at Rice University as part of the PIPER project. Producing a complete suite of performance tools for exascale platforms during the course of this project was impossible since both hardware and software for exascale systems is still a moving target. For that reason, the project focused broadly on the development of new techniques for measurement and analysis of performance on modern parallel architectures, enhancements to HPCToolkit’s software infrastructure to support our research goals or use on sophisticated applications, engaging developers of multithreaded runtimes to explore how support for tools should be integrated into their designs, engaging operating system developers with feature requests for enhanced monitoring support, engaging vendors with requests that they add hardware measure- ment capabilities and software interfaces needed by tools as they design new components of HPC platforms including processors, accelerators and networks, and finally collaborations with partners interested in using HPCToolkit to analyze and tune scalable parallel applications.

97 MATHEMATICS AND COMPUTING↗

ColDICE: A parallel Vlasov–Poisson solver using moving adaptive simplicial tessellation

Resolving numerically Vlasov–Poisson equations for initially cold systems can be reduced to following the evolution of a three-dimensional sheet evolving in six-dimensional phase-space. We describe a public parallel numerical algorithm consisting in representing the phase-space sheet with a conforming, self-adaptive simplicial tessellation of which the vertices follow the Lagrangian equations of motion. The algorithm is implemented both in six- and four-dimensional phase-space. Refinement of the tessellation mesh is performed using the bisection method and a local representation of the phase-space sheet at second order relying on additional tracers created when needed at runtime. In order to preserve in the best way the Hamiltonian nature of the system, refinement is anisotropic and constrained by measurements of local Poincaré invariants. Resolution of Poisson equation is performed using the fast Fourier method on a regular rectangular grid, similarly to particle in cells codes. To compute the density projected onto this grid, the intersection of the tessellation and the grid is calculated using the method of Franklin and Kankanhalli [65–67] generalised to linear order. As preliminary tests of the code, we study in four dimensional phase-space the evolution of an initially small patch in a chaotic potential and the cosmological collapse of a fluctuation composed of two sinusoidal waves. We also perform a “warm” dark matter simulation in six-dimensional phase-space that we use to check the parallel scaling of the code.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Time Series Classification for Locating Forced Oscillation Sources

Here, this article presents a machine learning based time-series classification method for using synchrophasor measurements to locate the source of forced oscillation (FO) for fast disturbance removal. First, multivariate time series (MTS) matrices are constructed by the most informative measurements selected by sequential feature selection from each power plant. Then, the Mahalanobis matrix is trained such that the Mahalanobis distance between the MTSs from the same class (i.e., with the same FO source location) are minimized and from different classes (i.e., with different FO source locations) are maximized. This allows MTSs to be classified by classifiers with class membership corresponding to the location of each FO source. To meet the runtime requirements of online matching, class templates are constructed to reduce data size and improve matching efficiency. To account for uncertainty in identifying the exact beginning of an FO event, dynamic time warping is used to align the out-of-sync MTSs. IEEE 39bus and WECC 179bus systems are used for algorithm development and validation. Simulation results demonstrate that the algorithm meets online operation runtime requirement with high accuracy using misaligned data sets.

42 ENGINEERING↗

IMEX and exact sequence discretization of the multi-fluid plasma model

Multi-fluid plasma models, where an electron fluid is modeled in addition to multiple ion and neutral species as well as the full set of Maxwell's equations, are useful for representing physics beyond the scope of classic MHD. This advantage presents challenges in appropriately dealing with electron dynamics and electromagnetic behavior characterized by the plasma and cyclotron frequencies and the speed of light. For physical systems, such as those near the MHD asymptotic regime, this requirement drastically increases runtimes for explicit time integration even though resolving fast dynamics may not be critical for accuracy. Implicit time integration methods, with efficient solvers, can help to step over fast time-scales that constrain stability, but do not strongly influence accuracy. As an extension, Implicit-explicit (IMEX) schemes provide an additional mechanism to choose which dynamics are evolved using an expensive implicit solve or resolved using a fast explicit solve. In this study, in addition to IMEX methods we also consider a physics compatible exact sequence spatial discretization. Here, this combines nodal bases (H-Grad) for fluid dynamics with a set of vector bases (H-Curl and H-Div) for Maxwell's equations. This discretization allows for multi-fluid plasma modeling without violating Gauss' laws for the electric and magnetic fields. This initial study presents a discussion of the major elements of this formulation and focuses on demonstrating accuracy in the linear wave regime and in the MHD limit for both a visco-resistive and a dispersive ideal MHD problem.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Bootstrap embedding for interacting electrons in phonon coherent-state mean field

Here, we develop a Fermi–Bose bootstrap embedding framework for the ground state of interacting electrons coupled to a phonon mean field. The method combines bootstrap embedding for correlated electrons with a self-consistent coherent-state mean-field treatment for phonons. This method models the interacting electron–phonon problem as a system of correlated electrons traveling in a self-consistently specified potential landscape, allowing for efficient treatment of large lattice systems. Convergence of the methods for fragment size and total system size is demonstrated for the one-dimensional Hubbard–Holstein model for up to 350 sites. Finite-size scaling is performed to extrapolate to the infinite system size. Benchmarking against the density matrix renormalization group for a small 8-site system at half- and quarter-filling shows an orders-of-magnitude runtime advantage. The comparison further reveals that the method performs best in regimes dominated by localization, such as the Mott insulating phase and the strong-coupling tiny polaron regime, where the local embedding ansatz is still valid. However, due to the mean-field treatment for phonons, we find limitations of our methods in the weakly coupled delocalized region and at the Peierls transition, where quantum phonon fluctuations and long-range kinetic correlations become substantial.

Islam, Shariful [North Carolina State University, ↗

Distributed Quantum Learning with co-Management in a Multi-tenant Quantum System

The rapid advancement of quantum computing has pushed classical designs into the quantum domain, breaking physical boundaries for computing-intensive and data-hungry applications with the hope that some systems may provide a quantum speedup. For example, variational quantum algorithms have been proposed for quantum neural networks to train deep learning models on qubits, achieving promising results. Existing quantum learning architectures and systems rely on single, monolithic quantum machines with abundant and stable resources, such as qubits. However, fabricating a large, monolithic quantum device is considerably more challenging than producing an array of smaller devices. In this paper, we investigate a distributed quantum system that combines multiple quantum machines into a unified system. We propose DQuLearn, which divides a quantum learning task into multiple subtasks. Each subtask can be executed distributively on individual quantum machines, with the results looping back to classical machines for subsequent training iterations. Additionally, our system supports multiple concurrent clients and dynamically manages their circuits according to the runtime status of quantum workers. Through extensive experiments, we demonstrate that DQuLearn achieves similar accuracies with significant runtime reduction, by up to 68.7% and an increase per-second circuit processing speed, by up to 3.99 times, in a 4-worker multi-tenant setting.

quantum computing↗

Parallel Application Performance on Two Generations of Intel Xeon HPC Platforms

Two next-generation node configurations hosting the Haswell microarchitecture were tested with a suite of microbenchmarks and application examples, and compared with a current Ivy Bridge production node on NREL" tm s Peregrine high-performance computing cluster. A primary conclusion from this study is that the additional cores are of little value to individual task performance--limitations to application parallelism, or resource contention among concurrently running but independent tasks, limits effective utilization of these added cores. Hyperthreading generally impacts throughput negatively, but can improve performance in the absence of detailed attention to runtime workflow configuration. The observations offer some guidance to procurement of future HPC systems at NREL. First, raw core count must be balanced with available resources, particularly memory bandwidth. Balance-of-system will determine value more than processor capability alone. Second, hyperthreading continues to be largely irrelevant to the workloads that are commonly seen, and were tested here, at NREL. Finally, perhaps the most impactful enhancement to productivity might occur through enabling multiple concurrent jobs per node. Given the right type and size of workload, more may be achieved by doing many slow things at once, than fast things in order.

97 MATHEMATICS AND COMPUTING↗

Vitrification of Hanford Tank 241-AP-107 with Recycled Condensate

During the vitrification of nuclear waste at the Hanford Waste Treatment and Immobilization Plant (WTP) – the primary mission of the U.S. Department of Energy Office of River Protection – the offgas condensate generated from the waste-to-glass conversion is currently planned to be concentrated by evaporation in the Effluent Management Facility (EMF). This concentrated condensate can then be recycled back to the incoming waste and vitrified. To test the recycle process, a test apparatus was designed to mimic the EMF evaporator and used to concentrate a volume of condensate that had been previously produced during the vitrification of Hanford tank 241-AP-107 (referred to herein as AP-107) waste in a continuous laboratory-scale melter (CLSM). The test apparatus successfully concentrated the AP-107 condensate by a factor of 10 while retaining over 90 % of the technetium-99 ( 99 Tc), Cs, and I inventory. A second portion of AP-107 waste was retrieved by Washington River Protection Solutions, LLC, given to Pacific Northwest National Laboratory, and combined with the AP-107 condensate concentrate after undergoing solids filtration and cesium removal by ion exchange. This combination served to approximate the recycling action to be performed at the WTP. After the addition of glass-forming chemicals (GFCs), the combined AP-107 waste and AP-107 condensate concentrate were processed in the CLSM to produce a glass, called AP-107-1R, that was designed to satisfy the WTP baseline requirements (Kim et al. 2012). During the 8.87 hours of processing, 7.27 kg of AP-107-1R glass were produced for an average glass production rate of 1739 kg m 2 d -1 . Compared to the previous run in the CLSM without recycled condensate, the run with the recycle had a greater average glass production rate, but the rate was within the potential range of variability when processing melter feeds with similar composition in the CLSM. The glass produced from the AP-107 recycle run in the CLSM was within 10 % of the target AP-107-1R glass composition with respect to the primary glass components. Analysis of the minor component impurities revealed that their content in the glass product had approached their nominal target after 2 turnovers of the glass inventory in the CLSM while the activity of the minor radionuclides was retained in the glass product. The 99 Tc and total cesium content in the combined AP-107 waste and recycled condensate were maintained at concentrations expected to be experienced at the WTP. During processing in the CLSM, at discrete sampling time periods, the target 99 Tc/Cs mass ratio in the glass formulation varied from 0.9 to 62.9. Across this range, the Cs retention in the glass ranged from 53 to 60 %, while the retention from the entire runtime totaled 68 %, values which align with Cs retention in other scaled melter systems while processing LAW melter feeds at 99 Tc/Cs mass ratios ranging from 1 to 100. The 99 Tc retention in the glass ranged from 22 to 32 %, primarily due to the cold-cap coverage on the glass melt surface, the area covered by reacting melter feed, varying from ~80 % to ~95 % during processing, demonstrating greater volatility of 99 Tc from the glass while more surface was exposed, as expected based on previous 99 Tc volatility studies.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

FleCSI Intro 2.0 [Slides]

FleCSI is a programming system or framework for developing mulitphysics application codes. FleCSI provides two primary capabilities--runtime abstraction layer and topology data structures.

97 MATHEMATICS AND COMPUTING↗

Adaptive time stepping for the two-time integro-differential Kadanoff-Baym equations

The nonequilibrium Green's function gives access to one-body observables for quantum systems. Of particular interest are quantities such as density, currents, and absorption spectra which are important for interpreting experimental results in quantum transport and spectroscopy. We present an integration scheme for the Green's function's equations of motion, the Kadanoff-Baym equations (KBE), which is both adaptive in the time integrator step size and method order as well as the history integration order. We analyze the importance of solving the KBE self-consistently and show that adapting the order of history integral evaluation is important for obtaining accurate results. To examine the efficiency of our method, we compare runtimes to a state-of-the-art fixed time step integrator for several test systems and show an order of magnitude speedup at similar levels of accuracy. Published by the American Physical Society 2024

97 MATHEMATICS AND COMPUTING↗