Search NASA⌕ Search

SEARCH · Search NASA

Results for “cloud connected”

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 541 records · Page 30

Interstellar organics and possible connections with the carbonaceous components of meteorites and IDPs

Studying the chemical and isotopic composition of interstellar ice and dust provides insight into the composition and chemical history of the solid bodies in the solar nebula and the nature of the materials subsequently brought into the inner part of the solar system by comets and meteorites. It is now possible to probe the composition of these microscopic interstellar particles. High quality IR spectra of many different astronomical sources (some associated with dark molecular clouds, and others in the diffuse interstellar medium) are now available. Comparisons of these spectra with laboratory spectra and complex organic components of these materials are discussed.

Allamandola, L. J.↗

Final Technical Report: Results of Phase 1

Arizona Public Service Company (APS) expects that by 2027, renewable energy will account for 6,590 GWh in energy consumption by its customers. While much of this future energy will come from large centrally-located power plants, distributed renewable energy, sited at the point of end-use will also play an important role in meeting the needs of APS customers and is expected to provide 1,734 GWh. With increasing penetration of residential and commercial photovoltaic (PV) systems at the point of end-use, PV power generation not only offsets the load, but could also cause significant shifts in power flow patterns through the distribution system, and could possibly cause reversal of flow through some branches of a distribution circuit. Significant changes to power flow introduced into existing distribution systems due to the increased amount of PV systems may cause operational issues, including over-voltage on the distribution feeder (loss of voltage regulation) and incorrect operation of control equipment, which may lead to an increase in the number of operations and related equipment wear that could affect equipment reliability and customer power quality. Additionally, connecting generation resources to a distribution feeder can introduce additional sources of short-circuit current to the distribution system. This could potentially result in increased short-circuit currents, potentially reaching damaging levels, causing protection desensitization and a potential loss of protection coordination. These effects may be further compounded by variability of PV production due to shading by clouds. The effects of these phenomena in distributed PV applications are not well understood, and there is a great need to characterize this variability. This project will contribute to understanding the effects of high-penetration solar electricity on the design and operation of distribution systems by demonstrating how a high penetration of PV systems affects grid operations of a working, utility distribution feeder. To address the technical challenges related to the integration of distributed PV when PV penetration levels reach or exceed 30% of the total load, technologies and methods to ensure the stable and safe operation of the feeder will be evaluated. Lessons learned will enable APS to improve the framework for future PV integration on its system and may also aid other utilities across the United States energy sector in accelerating the adoption of distributed photovoltaic generation.

14 SOLAR ENERGY↗

Further Examinations of Bimodal SLD Ice Accretion in the NASA Icing Research Tunnel

An ice shape database has been created to document ice accretions on a 72-inch chord NACA 23012 airfoil model resulting from an exposure to a Supercooled Large Drop (SLD) icing cloud with a bimodal drop size distribution. The ice shapes created were documented with photographs, laser scanned surface measurements over a section of the model span, and measurement of the ice mass over the same section of each accretion. The icing conditions used in the test matrix were based upon previously measured ice shapes on the same model to connect the current database to previously measured information. Ice shapes resulting from the bimodal distribution as well as from equivalent mono-modal drop size distributions were obtained and compared. Results indicate that the ice shapes resulting from the bimodal drop size distributions had higher mass and volume values than their mono-modal distribution equivalents as well as having icing limits that extended further back on the chord of the model.

Aircraft↗

Scanning Mobility Particle Sizer-Portable Optical Particle Spectrometer Intercomparison Field Campaign Report

A portable optical particle spectrometer (POPS) has recently been acquired by the United Kingdom Meteorological (UK Met) Office with the intention of mounting it on an unmanned aerial vehicle (UAV) for atmospheric aerosol research. Before conducting research with the new instrument, an intercomparison has been conducted to test the performance of the POPS against a scanning mobility particle sizer (SMPS) — a veteran and well-characterized particle sizer. As part of the Cloud-Aerosol-Radiation Interactions and Forcing (CLARIFY) campaign, the POPS was installed at the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Climate Research Facility mobile site on Ascension Island in the South Atlantic Ocean, alongside an ARM-operated SMPS. Both instruments sampled continuously from 20 August to 9 September, 2017. The SMPS and the POPS were connected to a common aerosol inlet; however, in the case of the SMPS, the sample air was dried before it entered the instrument. The POPS is a miniaturized optical particle counter designed specifically for use in balloon and UAV applications. The instrument is fully described in Goa et al. (2016) and so is only briefly described here. The POPS samples particles by drawing air through an inlet tube into an optical chamber, where it is illuminated by a 405nm laser. A sheath air flow is used to focus the sample air into the center of the laser beam, and the sample flow is maintained at a near-constant rate by an onboard pump. Side-scattered laser light is reflected into a photomultiplier tube by a hemispherical mirror, and the signal amplitude recorded by a data logger. Individual particle sizes are then inferred by comparing the recorded signal amplitudes to scattering amplitudes calculated using Mie theory. In common with other optical particle counters, the POPS size distributions are influenced by the refractive index used in the Mie calculations. The POPS is calibrated by the manufacturers (Handix Scientific) using latex spheres, with a refractive index (RI) of 1.615+0.001i. Sampling aerosols with a different refractive index, particularly if they are strongly absorbing, will result in significant uncertainties/errors in the particle size distributions (PSDs). We have therefore used scattering amplitude curves (calculated by Handix) for more absorbing aerosols, with an RI of 1.54+0.027i, which is expected to be more representative of the biomass-burning aerosol particles sampled at the ARM site during the CLARIFY campaign. In contrast to the POPS, the SMPS measures particle size using the differential mobility technique after applying an electrical charge to aerosol particles. The method is independent of the refractive index (Ruzer 2013).

CLARIFY, portable optical particle spectrometer, s↗

Assessing Customer Experience and Business Models around Price-to-Device Communication and Smart Control Pathways in CalFlexHub

California is facing three major challenges in electrical grid operation: renewable overgeneration, steep evening ramping, and growing peak demand. The state has identified dynamic retail price response as a key strategy evidenced by CPUC’s Dynamic Rates proceeding and CEC’s Load Management Standards. Furthermore, the CEC launched a $16M “California Load Flexibility Research and Deployment Hub (CalFlexHub)” administered by Berkeley Lab to accelerate price-response flexible load technologies in buildings and EV charging. There are more than 16 laboratory and field demonstration projects in CalFlexHub, each demonstrating innovative automated price-response technologies. CalFlexHub tests various pathways through which hourly price signals and triggered control commands are communicated to load-flexible devices such as smart thermostats, heat pumps, water heaters, and EVs. We identified seven unique communication and control pathways, which involve combinations of third-party cloud, device OEM’s cloud, building central gateway, and local controller in between the price server and the load-flexible devices. It is important for utilities and policy makers to understand the long-term implications of each pathway in designing future programs and creating related policies and mandates for market transformation. We propose an evaluation framework including the following aspects: ● Functionality: connectivity and uptime, resilience, and optimization; ● Customer experience: simplicity in setup, troubleshooting support, continuity, customer choice, first cost, and ongoing cost; ● Business model and scalability: advance interoperability, holistic solution, bridge unique gap, customer base, and value streams and pricing structures. In this paper, we identify emerging business models associated with each communication pathway and discuss their positive features and challenges from the above aspects.

Liu, Jingjing↗

The Large-Eddy Simulation (LES) Atmospheric Radiation Measurement (ARM) Symbiotic Simulation and Observation (LASSO) Activity for Continental Shallow Convection

The U.S. Department of Energy’s Atmospheric Radiation Measurement (ARM) user facility recently initiated the Large-Eddy Simulation (LES) ARM Symbiotic Simulation and Observation (LASSO) activity focused on shallow convection at ARM’s Southern Great Plains (SGP) atmospheric observatory in Oklahoma. LASSO is designed to overcome an oft-shared difficulty of bridging the gap from point-based measurements to scales relevant for model parameterization development, and it provides an approach to add value to observations through modeling. LASSO is envisioned to be useful to modelers, theoreticians, and observationalists needing information relevant to cloud processes. LASSO does so by combining a suite of observations, LES inputs and outputs, diagnostics, and skill scores into data bundles that are freely available, and by simplifying user access to the data to speed scientific inquiry. The combination of relevant observations with observationally constrained LES output provides detail that gives context to the observations by showing physically consistent connections between processes based on the simulated state. A unique approach for LASSO is the generation of a library of cases for days with shallow convection combined with an ensemble of LES for each case. The library enables researchers to move beyond the single-case-study approach typical of LES research. The ensemble members are produced using a selection of different large-scale forcing sources and spatial scales. Since large-scale forcing is one of the most uncertain aspects of generating the LES, the ensemble informs users about potential uncertainty for each date and increases the probability of having an accurate forcing for each case.

54 ENVIRONMENTAL SCIENCES↗

Automatic point Cloud Building Envelope Segmentation (Auto-CuBES) using Machine Learning

Modern retrofit construction practices use 3D point cloud data of the building envelope to obtain the as-built dimensions. However, manual segmentation by a trained professional is required to identify and measure window openings, door openings, and other architectural features, making the use of 3D point clouds labor-intensive. In this study, the Automatic point Cloud Building Envelope Segmentation (Auto-CuBES) algorithm is described, which can significantly reduce the time spent during point cloud segmentation. The Auto-CuBES algorithm inputs a 3D point cloud generated by commonly available surveying equipment and outputs a wire-frame model of the building envelope. Unsupervised machine learning methods were used to identify facades, windows, and doors while minimizing the number of calibration parameters. Additionally, Auto-CuBES generates a heat map of each facade indicating non-planar characteristics that are crucial for the optimization of connections used in overclad envelope retrofits. With a scan resolution of 3 mm, the resulting window dimensions showed a mean absolute error of 4.2 mm compared to manual laser measurements.

Maldonado Puente, Bryan↗

CHEMICAL ABUNDANCE ANALYSIS OF MOVING GROUP W11450 (LATHAM 1)

We present elemental abundances for all seven stars in Moving Group W11450 (Latham 1) to determine if they may be chemically related. These stars appear to be both spatially and kinematically related, but no spectroscopic abundance analysis exists in literature. Abundances for eight elements were derived via equivalent width analyses of high-resolution (R ∼ 60,000), high-signal-to-noise ratio (〈S/N〉∼100) spectra obtained with the Otto Struve 2.1 m telescope and the Sandiford Echelle Spectrograph at McDonald Observatory. The large star-to-star scatter in metallicity, −0.55 ≤ [Fe/H] ⩽ 0.06 dex (σ = 0.25), implies these stars were not produced from the same chemically homogeneous molecular cloud, and are therefore not part of a remnant or open cluster as previously proposed. Prior to this analysis, it was suggested that two stars in the group, W11449 and W11450, are possible wide binaries. The candidate wide binary pair show similar chemical abundance patterns with not only iron but with other elements analyzed in this study, suggesting the proposed connection between these two stars may be real.

79 ASTRONOMY AND ASTROPHYSICS↗

High Resolution Nature Runs and the Big Data Challenge

NASA's Global Modeling and Assimilation Office at Goddard Space Flight Center is undertaking a series of very computationally intensive Nature Runs and a downscaled reanalysis. The nature runs use the GEOS-5 as an Atmospheric General Circulation Model (AGCM) while the reanalysis uses the GEOS-5 in Data Assimilation mode. This paper will present computational challenges from three runs, two of which are AGCM and one is downscaled reanalysis using the full DAS. The nature runs will be completed at two surface grid resolutions, 7 and 3 kilometers and 72 vertical levels. The 7 km run spanned 2 years (2005-2006) and produced 4 PB of data while the 3 km run will span one year and generate 4 BP of data. The downscaled reanalysis (MERRA-II Modern-Era Reanalysis for Research and Applications) will cover 15 years and generate 1 PB of data. Our efforts to address the big data challenges of climate science, we are moving toward a notion of Climate Analytics-as-a-Service (CAaaS), a specialization of the concept of business process-as-a-service that is an evolving extension of IaaS, PaaS, and SaaS enabled by cloud computing. In this presentation, we will describe two projects that demonstrate this shift. MERRA Analytic Services (MERRA/AS) is an example of cloud-enabled CAaaS. MERRA/AS enables MapReduce analytics over MERRA reanalysis data collection by bringing together the high-performance computing, scalable data management, and a domain-specific climate data services API. NASA's High-Performance Science Cloud (HPSC) is an example of the type of compute-storage fabric required to support CAaaS. The HPSC comprises a high speed Infinib and network, high performance file systems and object storage, and a virtual system environments specific for data intensive, science applications. These technologies are providing a new tier in the data and analytic services stack that helps connect earthbound, enterprise-level data and computational resources to new customers and new mobility-driven applications and modes of work. In our experience, CAaaS lowers the barriers and risk to organizational change, fosters innovation and experimentation, and provides the agility required to meet our customers' increasing and changing needs

big data analysis↗

Datum: A Scientific Metadata Catalog

The data catalog market is currently flooded with a myriad of different products, but none serve the scientific community well. There are cloud-native tools like Databricks, Snowflake,to on-premise solutions like Collibra and Datahub. The common failing of all these tools however, is their inability to serve the scientific data community directly. Most catalogs are targeted towards financial, health, or user data - not sensor or scientific domain data. They also prioritize integrations that often don’t exist or are just starting to be used in the scientific realm - all while ignoring common scientific tools and file types. Datum is a catalog which targets the scientific data directly, including the tools and networks in which those tools are used. We work with the producers and consumers of the data where they are, targeting cloud and on-premise with a focus on classified networks. Datum is an Erlang/Elixir application. Technical Features Note: The features listed below are still under development and may change, slightly, upon final delivery of the product. File Formats - Datum has the ability to read additional metadata and provides processing pipelines for the following file formats: Plain Text, PDF, LaTeX, HTML, Open Document Format (.odt), XML, CSV/TSV (and other standard delimiters), OpenDocument Database and Spreadsheets, Geo-Referenced TIFF, Common Data Format, HDF/HDF5, LabView TDMS, Excel, DeltaTables, Parquet, Apache Iceberg, Apache Hudi and many others. Metadata Collection - Scanners for the local and networked file systems and cloud storage providers. Network integration with common databases such as MSSQL and MySQL. User Plugin System - Users are able to provide either file processing, metadata extraction, or sampling plugins in the programming language of their choice. Authentication/Authorization -: OIDC integration, SCIM provisioning and EntraID integration out of the box. Full user and group management system with a “least privilege” operating mode. Governance - Customizable data governance platform; dictate and enforce required metadata, enforce data embargos, and enforce user agreements and NDAs before data access. Ability to create health checks on data, rejecting abandoned or poorly curated data and automatically removing it from the search index. Ability for users to submit corrections. Search - Semantic search is a first class citizen. No licenses to expensive, external software required. Integrated use of vectors and vector-based search allows for AI agent integration at all levels of operation. Metadata Model - Display and control data’s lineage and connections to other data and data directories. Data is modeled after a filesystem - an organization instantly recognizable and navigable by most any user. CLI and SDK - Ships with a Command Line Interface (CLI) tool and with a fully-featured Python SDK. This allows for rapid and programmatic use of Datum by every level of user. Minimal Infrastructure - Datum ships as a single executable file and can be run on any operating system and most CPU architectures. Datum has no reliance on external databases, search indexing tools, or other outside services - and it runs equally well on edge computing devices, cloud services, or in a clustered HPC environment.

darrington, john↗

Identifying and tracking bubbles and drops in simulations: A toolbox for obtaining sizes, lineages, and breakup and coalescence statistics

Knowledge of bubble and drop size distributions in two-phase flows is important for characterizing a wide range of phenomena, including combustor ignition, sonar communication, and cloud formation. The physical mechanisms driving the background flow also drive the time evolution of these distributions. Accurate and robust identification and tracking algorithms for the dispersed phase are necessary to reliably measure this evolution and thereby quantify the underlying mechanisms in interface-resolving flow simulations. The identification of individual bubbles and drops traditionally relies on an algorithm used to identify connected regions. This traditional algorithm can be sensitive to the presence of spurious structures. A cost-effective refinement is proposed to maximize volume accuracy while minimizing the identification of spurious bubbles and drops. An accurate identification scheme is crucial for distinguishing bubble and drop pairs with large size ratios. The identified bubbles and drops need to be tracked in time to obtain breakup and coalescence statistics that characterize the evolution of the size distribution, including breakup and coalescence frequencies, and the probability distributions of parent and child bubble and drop sizes. An algorithm based on mass conservation is proposed to construct bubble and drop lineages using simulation snapshots that are not necessarily from consecutive time steps. These lineages are then used to detect breakup and coalescence events, and obtain the desired statistics. Accurate identification of large-size-ratio bubble and drop pairs enables accurate detection of breakup and coalescence events over a large size range. Accurate detection of successive breakup and coalescence events requires that the snapshot interval be an order of magnitude smaller than the characteristic breakup and coalescence times to capture these successive events while minimizing the identification of repeated confounding events. Together, these algorithms serve as a toolbox for detailed analysis of two-phase simulations, and enable insights into the mechanisms behind bubble and drop formation and evolution in flows of practical importance.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The CO Emission in the Taffy Galaxies (UGC 12914/15) at 60 pc Resolution. I. The Battle for Star Formation in the Turbulent Taffy Bridge

We present Atacama Large Millimeter/submillimeter Array observations at a spatial resolution of $0^{"}_{.}2$ (60 pc) of CO emission from the Taffy galaxies (UGC 12914/5). The observations are compared with narrowband Paα, mid-IR, radio continuum and X-ray imaging, plus optical spectroscopy. The galaxies have undergone a recent head-on collision, creating a massive gaseous bridge that is known to be highly turbulent. The bridge contains a complex web of narrow molecular filaments and clumps. The majority of the filaments are devoid of star formation, and fall significantly below the Kennicutt–Schmidt relationship for normal galaxies, especially for the numerous regions undetected in Paα emission. Within the loosely connected filaments and clumps of gas we find regions of high velocity dispersion that appear gravitationally unbound for a wide range of likely values of X CO . Like the "Firecracker" region in the Antennae system, they would require extremely high external dynamical or thermal pressure to stop them dissipating rapidly on short crossing timescales of 2–5 Myr. We suggest that the clouds may be transient structures within a highly turbulent multiphase medium that is strongly suppressing star formation. Despite the overall turbulence in the system, stars seem to have formed in compact hotspots within a kiloparsec-sized extragalactic H ii region, where the molecular gas has a lower velocity dispersion than elsewhere, and shows evidence for a collision with an ionized gas cloud. Like the shocked gas in the Stephan's Quintet group, the conditions in the Taffy bridge shows how difficult it is to form stars within a turbulent, multiphase, gas.

79 ASTRONOMY AND ASTROPHYSICS↗

Improving Access to MODIS Biophysical Science Products for NACP Investigators

MODIS 4 NACP is a NASA-funded project supporting the North American Carbon Program (NACP). The purpose of this Advancing Collaborative Connections for Earth-Sun System Science (ACCESS) project is to provide researchers with Moderate Resolution Imaging Spectroradiometer (MODIS) biophysical data products that are custom tailored for use in NACP model studies. Standard MODIS biophysical products provide used to improve our understanding on the climate and ecosystem changes. However, direct uses of the MODIS biophysical parameters are constrained by retrieval quality and cloud contamination. Another challenge that NACP users face is acquiring MODIS data in formats and at spatial-temporal resolutions consistent with other data sets they use. We have been working closely with key NACP users to tailor the MODIS products to fit their needs. First, we provide new temporally smoothed and spatially continuous MODIS biophysical data sets. Second, we are distributing MODIS data at suitable spatial-temporal resolutions and in formats consistent with other data integration into model studies.

Wolfe, Robert E.↗

Spy in the GPU-box: Covert and Side Channel Attacks on Multi-GPU System

The deep learning revolution has been enabled in large part by GPUs, and more recently accelerators, which make it possible to carry out computationally demanding training and inference in acceptable times. As the size of machine learning networks and workloads continues to increase, multi-GPU machines have emerged as an important platform offered on High Performance Computing and cloud data centers. Since these machines are shared among multiple users, it becomes increasingly important to protect applications against potential attacks. In this paper, we explore the vulnerability of Nvidia's DGX multi-GPU machines to covert and side channel attacks. These machines consist of a number of discrete GPUs that are interconnected through a combination of custom interconnect (NVLink) and PCIe connections. We reverse engineer the interconnected cache hierarchy and show that it is possible for an attacker on one GPU to cause contention on the L2 cache of another GPU. We use this observation to first develop a covert channel attack across two GPUs, achieving the best bandwidth of around 4 MB/s. We also develop a prime and probe attack on a remote GPU allowing an attacker to recover the cache access pattern of another workload. This access pattern can be used in any number of side channel attacks: we demonstrate a proof of concept attack that fingerprints the application running on the remote GPU, with high accuracy. We also develop a proof of concept attack to extract hyperparameters of a machine learning workload. Our work establishes for the first time the vulnerability of these machines to microarchitectural attacks and can guide future research to improve their security.

Dutta, Sankha↗

Interstellar Ice and Dust: The Feedstock of the Solar System

Studying the chemical and isotopic composition of interstellar ice and dust provides insight into the composition and chemical history of the solid bodies in the solar nebula and the nature of the material subsequently brought into the inner part of the solar system by comets and meteorites. It is now possible to probe the composition of these microscopic interstellar particles (some hundreds of light years away), thanks to substantial progress in two areas: astronomical spectroscopic techniques in the middle-infrared, the spectral region most diagnostic of composition; and laboratory simulations which realistically reproduce the critical conditions in various interstellar environments. High quality infrared spectra of many different astronomical sources, some associated with dark molecular clouds, and others in the diffuse interstellar medium (DISM) are now available. What comparisons of these spectra with laboratory spectra tell us about the complex organic components of these materials is the subject of this talk. Most interstellar material is concentrated in large molecular clouds where simple molecules are formed by gas phase and dust grain surface reactions. Gaseous species (except H2) striking the cold (10K) dust will stick, forming an icy grain mantle. This accretion, coupled with energetic particle bombardment and UV photolysis, will produce a complex chemical mixture containing volatile, non-volatile, and isotopically fractionated species. One can compare spectra of the diffuse and dense interstellar medium with the spectra of analogs produced in the laboratory under conditions which mimic those in these different environments. In this way one can determine the composition and abundances of the major constituents present and place general constraints on the types and relative abundances of organics coating the grains. Ices in dense clouds contain H2O, CH3OH, CO, perhaps some NH3 and H2CO, as well as nitriles and ketones or esters. There is some evidence that the later, more complex species, are also present on the grains in the DISM. The evidence for these materials, in addition to carbon rich materials such as amorphous carbon, microdiamonds, and polycyclic aromatic hydrocarbons will be reviewed and the possible connection with meteoritic organics will be discussed.

Allamandola, L. J.↗

Astronomical Problems and Laboratory Solutions

Studying the chemical and isotopic composition of interstellar ice and dust provides insight into the composition and chemical history of the solid bodies in the solar nebula and the nature of the material subsequently brought into the inner part of the solar system by comets and meteorites. It is now possible to probe the composition of these microscopic interstellar particles (some hundreds of light years away), thanks to substantial progress in two areas: astronomical spectroscopic techniques in the middle-infrared, the spectral region most diagnostic of composition; and laboratory simulations which realistically reproduce the critical conditions in various interstellar environments. High quality infrared spectra of many different astronomical sources, some associated with dark molecular clouds, and others in the diffuse interstellar medium (DISM) are now available. What comparisons of these spectra with laboratory spectra tell us about the complex organic components of these materials is the subject of this talk. Most interstellar material is concentrated in large molecular clouds where simple molecules are formed by gas phase and dust grain surface reactions. Gaseous species (except H2) striking the cold (10K) dust will stick, forming an icy grain mantle. This accretion, coupled with energetic particle bombardment and UV photolysis, will produce a complex chemical mixture containing volatile, non-volatile, and isotopically fractionated species. One can compare spectra of the diffuse and dense interstellar medium with the spectra of analogs produced in the laboratory under conditions which mimic those in these different environments. In this way one can determine the composition and abundances of the major constituents present and place general constraints on the types and relative abundances of organics coating the grains. Ices in dense clouds contain H2O, CH3OH, CO, perhaps some NH3 and H2CO, as well as nitriles and ketones or esters. there is some evidence that the later, more complex species, are also present on the grains in the DISM. The evidence for these materials, in addition to carbon rich materials such as amorphous carbon, microdiamonds, and polycyclic aromatic hydrocarbons will be reviewed and the possible connection with meteoritic organics will be discussed.

Allamandola, L. J.↗

Performance Profile of Transformer Fine-Tuning in Multi-GPU Cloud Environments

The study presented here focuses on performance characteristics and trade-offs associated with running machine-learning tasks in multi-GPU environments on both on-site cloud computing resources and commercial cloud services (Azure). Specifically, this study examines these tradeoffs by examining the performance of training and fine-tuning of transformer-based deep-learning (DL) networks on clinical notes and data, a task of critical importance in the medical domain. To this end, we perform DL-related experiments on the widely deployed NVIDIA V100 GPUs and on the newer A100 GPUs connected via NVLink or PCIe. This study analyzes the execution time of major operations to train DL models and investigate popular options to optimize each of them. We examine and present the findings on the impacts that various operations (e.g. data loading into GPUs, training, fine-tuning), optimizations, and system configurations (single vs. multi-GPU, NVLink vs. PCIe) have on the overall training performance.

Begoli, Edmon↗

On Lidar Sounding of the Atmosphere to Estimate Static and Dynamic Characteristics of Aerosol Inhomogeneities

1. A possible application of intensity fluctuations of a pulse light signal reflected by atmospheric aerosols is analyzed by the correlation method to evaluate static (medium sizes, shape) and dynamic (speed and direction of movement, lifetime) characteristics of aerosol inhomogeneities. The aerosol inhomogeneities are assumed to be expanded, pressed, disintegrated and originated constantly in accordance with random laws, the set of inhomogeneities as a whole traveling together with air masses and having predominant movement in wind direction. It is shown that the characteristics of aerosol inhomogeneities considered can be expressed by the coefficients of the correlation function expansion of the reflected signal fluctuation intensity in Tailor series. 2/ Correlation systems for evaluating static and dynamic characteristics of driving objects can be divided into two types according to the kind and quantity of used information: the systems with coordinates of the information removal "points" to be fixed in space, and the systems with a parallel simultaneous information removal at discrete moments of time. The systems for determination of wind direction considered in are the examples of the first type system. However, the operating information removal for two points is insufficient to estimate completely static and dynamic characteristics of inhomogenities, their quantity ought to be increased up to three of them for two-dimensional problem and up to four of them for three-dimensional problem as it is usually done in the ionospheric studies. The second type systems are used for the investigation of a medium shape and speed of the clouds according to photographs made from satellites. These systems are also used for solution of navigation problems. The use of optical quantum generators with a scanning beam is seen to increase greatly the working information removal in comparison with the first type systems. Nevertheless, scanning rate is not sufficient sometimes in order· to consider a general picture of aerosol inhomogeneities to be stationary. In this connection the use of the systems of second type treatment becomes a matter of essential difficulty. 3. Aerosol inhomogeneities simulation has been carried out on the basis of the digital computer experiments with the aim of estimating static and dynamic characteristics of inhomogeneities by an optical beam in the atmosphere at different scanning procedures. The dependence of determination accuracy of these characteristics on the type of chosen laws of aerosol particle distribution in the atmosphere, the parameters of inhomogeneities geometry, their speed and the law of scanning have been obtained.

Zuev, V. E.↗