Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed computation”

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 1,243 records · Page 69

Exploiting CRIC to streamline the configuration management of GlideinWMS factories for CMS support

GlideinWMS is a workload management and provisioning system that allows sharing computing resources distributed over independent sites. Based on the requests made by GlideinWMS frontends, a dynamically sized pool of resources is created by GlideinWMS pilot factories via pilot job submission to resource sites’ CEs. More than 400 CEs are currently serving more than ten virtual organizations through GlideinWMS, with CMS being the biggest user with 230 CEs. The complex configurations of the parameters defining resource requests, as submitted to those CEs, have been historically managed by manually editing a set of different XML files. New possibilities arise with CMS adopting the CRIC, an information system that collects, aggregates, stores, and exposes, among other things, computing resource data coming from various data providers. The paper will describe the challenges faced when CMS started to use CRIC to automatically generate the GlideinWMS factory configurations. The architecture of the prototype, and the ancillary tools developed to ease this transition, will be discussed. Finally, future plans and milestones will be outlined.

Dost, Jeffrey↗

Scaling Deep Learning on GPU and Knights Landing clusters

The speed of deep neural networks training has become a big bottleneck of deep learning research and development. For example, training GoogleNet by ImageNet dataset on one Nvidia K20 GPU needs 21 days. To speed up the training process, the current deep learning systems heavily rely on the hardware accelerators. However, these accelerators have limited on-chip memory compared with CPUs. To handle large datasets, they need to fetch data from either CPU memory or remote processors. We use both self-hosted Intel Knights Landing (KNL) clusters and multi-GPU clusters as our target platforms. From an algorithm aspect, current distributed machine learning systems are mainly designed for cloud systems. These methods are asynchronous because of the slow network and high fault-tolerance requirement on cloud systems. We focus on Elastic Averaging SGD (EASGD) to design algorithms for HPC clusters. Original EASGD used round-robin method for communication and updating. The communication is ordered by the machine rank ID, which is inefficient on HPC clusters. First, we redesign four efficient algorithms for HPC systems to improve EASGD's poor scaling on clusters. Async EASGD, Async MEASGD, and Hogwild EASGD are faster \textcolor{black}{than} their existing counterparts (Async SGD, Async MSGD, and Hogwild SGD, resp.) in all the comparisons. Finally, we design Sync EASGD, which ties for the best performance among all the methods while being deterministic. In addition to the algorithmic improvements, we use some system-algorithm codesign techniques to scale up the algorithms. By reducing the percentage of communication from 87% to 14%, our Sync EASGD achieves 5.3x speedup over original EASGD on the same platform. We get 91.5% weak scaling efficiency on 4253 KNL cores, which is higher than the state-of-the-art implementation.

60 APPLIED LIFE SCIENCES↗

Fusion neutron generation computations in a stellarator-mirror hybrid with neutral beam injection

In the paper [Moiseenko V.E., Noack K., Agren O. 'Stellarator-mirror based fusion driven fission reactor' J Fusion Energy 29 (2010) 65.], a version of a fusion driven system (FDS), i.e. a sub-critical fast fission assembly with a fusion plasma neutron source, is proposed. The plasma part of the reactor is based on a stellarator with a small mirror part. Hot ions with high perpendicular energy are assumed to be trapped in the magnetic mirror part. The stellarator part which connects to the mirror part and provides confinement for the bulk (deuterium) plasma. In the magnetic well of the mirror part, fusion reactions occur from collisions between a of hot ion component (tritium) with cold background plasma ions. RF heating is one option to heat the tritium. A more conventional method to sustain the hot ions is neutral beam injection (NBI), which is here studied numerically for the above-mentioned hybrid scheme. For these studies, a new kinetic code, KNBIM, has been developed. The code takes into account Coulomb collisions between the hot ions and the background plasma. The geometry of the confining magnetic field is arbitrary for the code. It is accounted for via a numerical bounce averaging procedure. Along with the kinetic calculations the neutron generation intensity and its spatial distribution are computed.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Properties and uncertainties of scalar field models of dark energy with barotropic equation of state

The dynamics of expansion and large scale structure formation in the multicomponent Universe with dark energy modeled by the minimally coupled scalar field with generalized linear barotropic equation of state are analyzed. It is shown that the past dynamics of expansion and future of the Universe - eternal accelerated expansion or turnaround and collapse - are completely defined by the current energy density of a scalar field and relation between its current and early equation of state parameters. The clustering properties of such models of dark energy and their imprints in the power spectrum of matter density perturbations depend on the same relation and, additionally, on the 'effective sound speed' of a scalar field, defined by its Lagrangian. It is concluded that such scalar fields with different values of these parameters are distinguishable in principle. This gives the possibility to constrain them by confronting the theoretical predictions with the corresponding observational data. For that we have used the 7-year Wilkinson Microwave Anisotropy Probe data on cosmic microwave background anisotropies, the Union2 data set on Supernovae Ia and the seventh data release of the Sloan Digital Sky Survey data on luminous red galaxies space distribution. Using the Markov Chain Monte Carlo technique the marginalized posterior and mean likelihood distributions are computed for the scalar fields with two different Lagrangians: Klein-Gordon and Dirac-Born-Infeld ones. The properties of such scalar field models of dark energy with best fitting parameters and uncertainties of their determination are also analyzed in the paper.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

iDDS: intelligent distributed dispatch and scheduling for workflow orchestration

The intelligent distributed dispatch and scheduling (iDDS) service is a versatile workflow orchestration system designed for large-scale, distributed scientific computing. iDDS extends traditional workload and data management by integrating data-aware execution, conditional logic, and programmable workflows, enabling automation of complex and dynamic processing pipelines. Originally developed for the ATLAS experiment at the large hadron collider, iDDS has evolved into an experiment-agnostic platform that supports both template-driven workflows and a Function-as-a-Task model for Python-based orchestration. This paper presents the architecture and core components of iDDS, highlighting its scalability, modular message-driven design, and integration with systems such as PanDA and Rucio. We demonstrate its versatility through real-world use cases: fine-grained tape resource optimization for ATLAS, orchestration of large Directed Acyclic Graph (DAG) workflows for the Rubin Observatory, distributed hyperparameter optimization for machine learning applications, active learning for physics analyses, and AI-assisted detector design at the electron–ion collider. By unifying workload scheduling, data movement, and adaptive decision-making, iDDS reduces operational overhead and enables reproducible, high-throughput workflows across heterogeneous infrastructures. We conclude with current challenges and future directions, including interactive, cloud-native, and serverless workflow support.

97 MATHEMATICS AND COMPUTING↗

CT dose equilibration and energy absorption in polyethylene cylinders with diameters from 6 to 55 cm

Purpose: ICRU Report No. 87 Committee and AAPM Task Group 200 designed a three-sectional polyethylene phantom of 30 cm in diameter and 60 cm in length for evaluating the midpoint dose D{sub L}(0) and its rise-to-the-equilibrium curve H(L) = D{sub L}(0)/D{sub eq} from computed tomography (CT) scanning, where D{sub eq} is the equilibrium dose. To aid the use of the phantom in radiation dose assessment and to gain an understanding of dose equilibration and energy absorption in polyethylene, the authors evaluated the short (20 cm) to long (60 cm) phantom dose ratio with a polyethylene diameter of 30 cm, assessed H(L) in polyethylene cylinders of 6–55 cm in diameters, and examined energy absorption in these cylinders. Methods: A GEANT4-based Monte Carlo program was used to simulate the single axial scans of polyethylene cylinders (diameters 6–55 cm and length 90 cm, as well as diameter 30 cm and lengths 20 and 60 cm) on a clinical CT scanner (Somatom Definition dual source CT, Siemens Healthcare). Axial dose distributions were computed on the phantom central and peripheral axes. An average dose over the central 23 or 100 mm region was evaluated for modeling dose measurement using a 0.6 cm{sup 3} thimble chamber or a 10 cm long pencil ion chamber, respectively. The short (20 cm) to long (90 cm) phantom dose ratios were calculated for the 30 cm diameter polyethylene phantoms scanned at four tube voltages (80–140 kV) and a range of beam apertures (1–25 cm). H(L) was evaluated using the dose integrals computed with the 90 cm long phantoms. The resultant H(L) data were subsequently used to compute the fraction of the total energy absorbed inside or outside the scan range (E{sub in}/E or E{sub out}/E) on the phantom central and peripheral axes, where E = LD{sub eq} was the total energy absorbed along the z axis. Results: The midpoint dose in the 60 cm long polyethylene phantom was equal to that in the 90 cm long polyethylene phantom. The short-to-long phantom dose ratios changed with beam aperture and phantom axis but were insensitive to tube voltage. H(L) was insensitive to tube voltage and CT scanner model. As phantom diameter increased from 6 to 55 cm, E{sub in}/E generally decreased but asymptotically approached constant levels on the peripheral axes of large phantoms. The curve of E{sub in}/E versus scan length was almost identical to that of H(L). Similarly, E{sub out}/E increased with scan length and asymptotically approached the equilibrium for large scan lengths. E{sub out}/D{sub eq} was much less than the equilibrium length L{sub eq} where H(L) = 0.98, even with scan lengths much larger than L{sub eq}. Conclusions: The polyethylene phantom designed by ICRU Report No. 87 Committee and AAPM Task Group 200 is adequately long for assessing the midpoint dose and its equilibration in CT scanning. The short-to-long phantom dose ratios and the H(L) data provided in this paper allow easy evaluations of the midpoint dose, longitudinal dose distribution, and energy absorption in polyethylene phantoms. The results of dose equilibration and energy absorption presented herein may be insightful for the clinical CT scans with various subject sizes and scan lengths.

07 ISOTOPE AND RADIATION SOURCES↗

Data-domain correlation approach for joint hydrogeologic inversion of time-lapse hydrogeologic and geophysical data

Inverse estimations of hydrogeologic properties are often highly uncertain due to the expense of collecting hydrogeologic data and the subsequent lack of information. Geophysical data can potentially help fill this information gap because geophysical methods can survey large areas remotely and relatively inexpensively. However, geophysical data are difficult to incorporate into hydrogeological parameter estimations due primarily to a lack of knowledge concerning the petrophysical relationships between hydrogeological and geophysical parameters. We present a method which allows time lapse geophysical data to be directly incorporated into a hydrogeological parameter estimation when there is a strong correlation between changes in geophysical and hydrogeological properties. This approach bypasses the need for an explicit petrophysical transform by formulating the geophysical part of the hydrogeological inversion in terms of a data domain correlation operator. We demonstrate the approach with a synthetic electrical resistivity monitoring application used to estimate the hydraulic conductivity distribution. Including time lapse resistivity data to supplement sparse hydrological data appears to greatly improve resolution of hydraulic conductivity in this case. More generally, the formulation and results suggest that geophysical monitoring data can be effectively incorporated into a hydrogeological parameter estimation using a data domain correlation operator, assuming there is a strong correlation between changes in hydrogeological and geophysical properties.

58 GEOSCIENCES↗

Iteration-based Linearized Distribution-level Locational Marginal Price for Three-phase Unbalanced Distribution Systems

Distributed energy resources (DERs) are rocking the utilities’ business landscape. It calls for competitive market environments that incentivize DERs to form maximum operating efficiency. Among proposed pricing schemes, distribution-level locational marginal price (DLMP) is effective in signaling the marginal generation cost differences driven by energy losses and network constraints. It can be derived from a distribution-level optimal power flow (OPF) framework, as it essentially presents the sensitivity of optimized generation cost towards incremental loads. However, due to the high resistance-to-inductance ratio and unbalanced characteristics of distribution networks, computational affordable DLMPs are highly challenged. This article provides a linear-approximated DLMP that can be solved efficiently and generalized to account for reactive power flow, three-phase unbalanced loads and meshed network structure. The successive linear programming technique is introduced to enhance the model accuracy. Case studies on an IEEE 123-Bus system validate its accuracy against a nonlinear benchmark and capability in offering proper incentives.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Generating Massive Scale-free Networks: Novel Parallel Algorithms using the Preferential Attachment Model

Recently, there has been substantial interest in the study of various random networks as mathematical models of complex systems. As real-life complex systems grow larger, the ability to generate progressively large random networks becomes all the more important. This motivates the need for efficient parallel algorithms for generating such networks. Naïve parallelization of sequential algorithms for generating random networks is inefficient due to inherent dependencies among the edges and the possibility of creating duplicate (parallel) edges. In this article, we present message passing interface-based distributed memory parallel algorithms for generating random scale-free networks using the preferential-attachment model. Our algorithms are experimentally verified to scale very well to a large number of processing elements (PEs), providing near-linear speedups. The algorithms have been exercised with regard to scale and speed to generate scale-free networks with one trillion edges in 6 minutes using 1,000 PEs.

97 MATHEMATICS AND COMPUTING↗

Large-Scale Optimization for Bayesian Inference in Complex Systems (Final Report)

The SAGUARO (Scalable Algorithms for Groundwater Uncertainty Analysis and Robust Optimization) Project focuses on the development of scalable numerical algorithms for large-scale Bayesian inversion in complex systems that capitalize on advances in large-scale simulation-based optimization and inversion methods. Our research is directed in three complementary areas: efficient approximations of the Hessian operator, reductions in complexity of forward simulations via stochastic spectral approximations and model reduction, and employing large-scale optimization concepts to accelerate sampling. Our efforts are integrated in the context of a challenging testbed problem that considers subsurface reacting flow and transport. The MIT component of the SAGUARO Project addresses the intractability of conventional sampling methods for large-scale statistical inverse problems by devising reduced-order models that are faithful to the full-order model over a wide range of parameter values; sampling then employs the reduced model rather than the full model, resulting in very large computational savings. Results indicate little effect on the computed posterior distribution. On the other hand, in the Texas-Georgia Tech component of the project, we retain the full-order model, but exploit inverse problem structure (adjoint-based gradients and partial Hessian information of the parameter-to- observation map) to implicitly extract lower dimensional information on the posterior distribution; this greatly speeds up sampling methods, so that fewer sampling points are needed. We can think of these two approaches as "reduce then sample" and "sample then reduce." In fact, these two approaches are complementary, and can be used in conjunction with each other. Moreover, they both exploit deterministic inverse problem structure, in the form of adjoint-based gradient and Hessian information of the underlying parameter-to-observation map, to achieve their speedups.

97 MATHEMATICS AND COMPUTING↗

Confronting dark energy models using galaxy cluster number counts

The mass function of cluster-size halos and their redshift distribution are computed for 12 distinct accelerating cosmological scenarios and confronted to the predictions of the conventional flat {Lambda}CDM model. The comparison with {Lambda}CDM is performed by a two-step process. First, we determine the free parameters of all models through a joint analysis involving the latest cosmological data, using supernovae type Ia, the cosmic microwave background shift parameter, and baryon acoustic oscillations. Apart from a braneworld inspired cosmology, it is found that the derived Hubble relation of the remaining models reproduces the {Lambda}CDM results approximately with the same degree of statistical confidence. Second, in order to attempt to distinguish the different dark energy models from the expectations of {Lambda}CDM, we analyze the predicted cluster-size halo redshift distribution on the basis of two future cluster surveys: (i) an X-ray survey based on the eROSITA satellite, and (ii) a Sunayev-Zeldovich survey based on the South Pole Telescope. As a result, we find that the predictions of 8 out of 12 dark energy models can be clearly distinguished from the {Lambda}CDM cosmology, while the predictions of 4 models are statistically equivalent to those of the {Lambda}CDM model, as far as the expected cluster mass function and redshift distribution are concerned. The present analysis suggests that such a technique appears to be very competitive to independent tests probing the late time evolution of the Universe and the associated dark energy effects.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Scattering of electromagnetic waves from two-dimensional randomly rough perfectly conducting surfaces: The full angular intensity distribution

By a computer simulation approach we study the scattering of p- or s-polarized light from a two-dimensional, randomly rough, perfectly conducting surface. The pair of coupled inhomogeneous integral equations for two independent tangential components of the magnetic field on the surface are converted into matrix equations by the method of moments, which are then solved by the biconjugate gradient stabilized method. The solutions are used to calculate the mean differential reflection coefficient for given angles of incidence and specified polarizations of the incident and scattered fields. The full angular distribution of the intensity of the scattered light is obtained for strongly randomly rough surfaces by a rigorous computer simulation approach.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Optimizing High-Throughput Inference on Graph Neural Networks at Shared Computing Facilities with the NVIDIA Triton Inference Server

Abstract With machine learning applications now spanning a variety of computational tasks, multi-user shared computing facilities are devoting a rapidly increasing proportion of their resources to such algorithms. Graph neural networks (GNNs), for example, have provided astounding improvements in extracting complex signatures from data and are now widely used in a variety of applications, such as particle jet classification in high energy physics (HEP). However, GNNs also come with an enormous computational penalty that requires the use of GPUs to maintain reasonable throughput. At shared computing facilities, such as those used by physicists at Fermi National Accelerator Laboratory (Fermilab), methodical resource allocation and high throughput at the many-user scale are key to ensuring that resources are being used as efficiently as possible. These facilities, however, primarily provide CPU-only nodes, which proves detrimental to time-to-insight and computational throughput for workflows that include machine learning inference. In this work, we describe how a shared computing facility can use the NVIDIA Triton Inference Server to optimize its resource allocation and computing structure, recovering high throughput while scaling out to multiple users by massively parallelizing their machine learning inference. To demonstrate the effectiveness of this system in a realistic multi-user environment, we use the Fermilab Elastic Analysis Facility augmented with the Triton Inference Server to provide scalable and high-throughput access to a HEP-specific GNN and report on the outcome.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

SU-E-T-94: An Advanced Rotating Gamma Ray System for Stereotactic Radiosurgery

Purpose: Co-60 beams have unique dosimetric properties that are ideally suited for cranial treatments. Co-60 sources with cone-shaped collimators provide conformal dose distributions allowing for ablative treatments with rapid dose falloff to spare nearby critical structures. This work investigates a novel, image-guided, rotational Gamma ray system that provides both superior dose conformity/gradient and accurate stereotaxy for stereotactic radiosurgery (SRS). Methods: The SupeRay system (Cyber Medical Corp., China) consists of a rotating source chamber containing 30 gamma sources focusing at the isocenter with 4 collimators measuring 3, 4, 8 and 16mm in diameter. A novel switch design enables the 30 Gamma sources to be turned off at any arbitrarily selected 60° interval in order to avoid critical structures. The 3D treatment couch provides automatic treatment positioning between individual shots and the kV imaging system provides orthogonal images with a spatial resolution of 0.24mm to facilitate target localization. Monte Carlo simulations were used to compute dose distributions and compare with measurements and other Gamma ray SRS systems. Results: Monte Carlo results confirmed the SupeRay design parameters including output factors and 3D dose distributions. Its beam penumbra/dose gradient is similar to or slightly better than that of the Elekta Gamma Knife. The penumbra in the (x,y,z) direction was (7.38mm,7.38mm,3.86mm) for the 16mm collimator, (4.83mm,4.83mm,3.12mm) for the 8mm collimator, and (3.03mm,3.03mm,2.38mm) for the 4mm collimator, respectively, on the SupeRay system while it was (9.5mm,10.0mm,2.9mm), (4.3mm,4.3mm,2.9mm) and (3.2mm,3.2mm,1.9mm) for the same collimator sizes, respectively, on the Perfexion system. The kV imaging system together with a non-invasive relocatable frame provides accurate target localization (<0.5mm) for cases requiring multiple treatment fractions. Conclusion: Because of the unique dosimetric properties of Co-60 sources and its accurate stereotaxy/dose delivery, SupeRay is ideally suited for single and multiple-session ablative treatments of cranial lesions. This work was partially supported by Cyber Medical Corp.

60 APPLIED LIFE SCIENCES↗

Large-Scale Optimization for Bayesian Inference in Complex Systems

The SAGUARO (Scalable Algorithms for Groundwater Uncertainty Analysis and Robust Optimization) Project focused on the development of scalable numerical algorithms for large-scale Bayesian inversion in complex systems that capitalize on advances in large-scale simulation-based optimization and inversion methods. The project was a collaborative effort among MIT, the University of Texas at Austin, Georgia Institute of Technology, and Sandia National Laboratories. The research was directed in three complementary areas: efficient approximations of the Hessian operator, reductions in complexity of forward simulations via stochastic spectral approximations and model reduction, and employing large-scale optimization concepts to accelerate sampling. The MIT₋Sandia component of the SAGUARO Project addressed the intractability of conventional sampling methods for large-scale statistical inverse problems by devising reduced-order models that are faithful to the full-order model over a wide range of parameter values; sampling then employs the reduced model rather than the full model, resulting in very large computational savings. Results indicate little effect on the computed posterior distribution. On the other hand, in the Texas₋Georgia Tech component of the project, we retain the full-order model, but exploit inverse problem structure (adjoint-based gradients and partial Hessian information of the parameter-to-observation map) to implicitly extract lower dimensional information on the posterior distribution; this greatly speeds up sampling methods, so that fewer sampling points are needed. We can think of these two approaches as "reduce then sample" and "sample then reduce." In fact, these two approaches are complementary, and can be used in conjunction with each other. Moreover, they both exploit deterministic inverse problem structure, in the form of adjoint-based gradient and Hessian information of the underlying parameter-to-observation map, to achieve their speedups.

97 MATHEMATICS AND COMPUTING↗

Covariant boost and structure functions of baryons in Gross-Neveu models

Baryons in the large N limit of two-dimensional Gross-Neveu models are reconsidered. The time-dependent Dirac-Hartree-Fock approach is used to boost a baryon to any inertial frame and shown to yield the covariant energy-momentum relation. Momentum distributions are computed exactly in arbitrary frames and used to interpolate between the rest frame and the infinite momentum frame, where they are related to structure functions. Effects from the Dirac sea depend sensitively on the occupation fraction of the valence level and the bare fermion mass and do not vanish at infinite momentum. In the case of the kink baryon, they even lead to divergent quark and antiquark structure functions at x=0.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

GridAPPS-D Conceptual Design v1.0

The purpose of this document is to provide a conceptual design of the distribution system application development platform being developed for the U.S. Department of Energy’s Advanced Distribution Management System (ADMS) Program by the Grid Modernization Laboratory Consortium project GM0063. The platform will be referred to as GridAPPS-D. This document provides a high level, conceptual view of the platform and provides related background and contextual information. This document is intended to both educate readers about the technical work of the project and to serve as a point of reference for the project team. The document will be updated as the project progresses.

97 MATHEMATICS AND COMPUTING↗