Search NASA⌕ Search

SEARCH · Search NASA

Results for “bottleneck detection”

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 73 records · Page 4

Enhancing Sensitivity in Targeted Single-Cell Proteomics by Coupling a Dual Ion Funnel Interface with Triple Quadrupole Mass Spectrometer

Single-cell proteomics (SCP) has emerged as a powerful approach for understanding cellular heterogeneity and biological processes at unprecedented resolution. However, the extremely limited protein content of individual cells (femtogram to picogram levels) pushes current mass spectrometry instrumentation to its sensitivity limits, creating a critical analytical bottleneck. While selected reaction monitoring (SRM) using triple quadrupole (QqQ) instruments 1 offers advantages in sensitivity and reproducibility for targeted proteomics quantification, SRM still struggles with sensitivity for quantification of moderate- or low-abundance proteins from single-cell sample amounts. Here, we report the development and systematic evaluation of a dual ion funnel interface designed to address the sensitivity limitation by significantly enhancing ion transmission efficiency in commercial QqQ mass spectrometers. The dual ion funnel interface, composed of a curved S-funnel followed by a conventional ion funnel, improves ion transmission efficiency while reducing chemical noise through selective ion focusing. The performance of the dual ion funnel interface was systematically compared to standard interface on a TSQ Vantage platform across samples with different levels of complexity. The dual funnel interface demonstrated to provide up to 25-fold improvement in sensitivity across a wide range of protein concentrations in different biological matrices (low complex mouse macrophage and high complex human cells). Critically, enhanced sensitivity was accompanied by increased analytical reproducibility with lower coefficient of variations. Most importantly, the dual funnel interface enabled reliable quantification of low-abundance proteins that were barely detectable or not detected by the standard interface, extending analysis to single-cell equivalent amounts while maintaining excellent reproducibility. These results demonstrate that the dual funnel interface addresses the critical bottleneck in quantitative targeted proteomics, providing a technological foundation for ultrasensitive targeted SCP that requires both high sensitivity and robust quantitative performance.

Min, Sehong↗

Large spin Hall effect in Si at room temperature

Silicon's weak intrinsic spin-orbit coupling and centrosymmetric crystal structure are a critical bottleneck to the development of Si spintronics, because they lead to an insignificant spin Hall effect (spin current generation) and inverse spin Hall effect (spin current detection). In this work, we undertake current, magnetic field, crystallography dependent magnetoresistance, and magnetothermal transport measurements to study the spin transport behavior in freestanding Si thin films. We observe a large spin Hall magnetoresistance in both $\mathcal{p}$ – Si and $\mathcal{n}$ – Si at room temperature and it is an order of magnitude larger than that of Pt. One explanation of the unexpectedly large and efficient spin Hall effect is spin-phonon coupling instead of spin-orbit coupling. The macroscopic origin of the spin-phonon coupling can be large strain gradients that can exist in the freestanding Si films. This discovery in a light, earth abundant and centrosymmetric material opens a new path of strain engineering to achieve spin dependent properties in technologically highly developed materials.

36 MATERIALS SCIENCE↗

Scalable Hyper-parameter Estimation for Gaussian Process Based Time Series Analysis

Gaussian process (GP) is increasingly becoming popular as a kernel machine learning tool for non-parametric data analysis. Recently, GP has been applied to model non-linear dependencies in time series data. GP based analysis can be used to solve problems of time series prediction, forecasting, missing data imputation, change point detection, anomaly detection, etc. But the use of GP to handle massive scientific time series data sets has been limited, owing to its expensive computational complexity. The primary bottleneck is the handling of the covariance matrix whose size is quadratic in the length of the time series. In this paper we propose a scalable method that exploit the special structure of the covariance matrix for hyper-parameter estimation in GP based learning. The proposed method allows estimation of hyper parameters associated with GP in quadratic time, which is an order of magnitude improvement over standard methods with cubic complexity. Moreover, the proposed method does not require explicit computation of the covariance matrix and hence has memory requirement linear to the length of the time series as opposed to the quadratic memory requirement of standard methods. To further improve the computational complexity of the proposed method, we provide a parallel version to concurrently estimate the log likelihood for a set of time series which is the key step in the hyper-parameter estimation. Performance results on a multi-core system show that our proposed method provides significant speedups as high as 1000, even when running in serial mode, while maintaining a small memory footprint. The parallel version exploits the natural parallelization potential of the serial algorithm and is shown to perform significantly better than the serial faster algorithm, with speedups as high as 10.

Chandola, Varun↗

Automated Grain Boundary (GB) Segmentation and Microstructural Analysis in 347H Stainless Steel Using Deep Learning and Multimodal Microscopy

Austenitic 347H stainless steel offers superior mechanical properties and corrosion resistance required for extreme operating conditions such as high temperature. The change in microstructure due to composition and process variations is expected to impact material properties. Identifying microstructural features such as grain boundaries thus becomes an important task in the process-microstructure-properties loop. Applying convolutional neural network (CNN)-based deep learning models is a powerful technique to detect features from material micrographs in an automated manner. In contrast to microstructural classification, supervised CNN models for segmentation tasks require pixel-wise annotation labels. However, manual labeling of the images for the segmentation task poses a major bottleneck for generating training data and labels in a reliable and reproducible way within a reasonable timeframe. Microstructural characterization especially needs to be expedited for faster material discovery by changing alloy compositions. Here, in this study, we attempt to overcome such limitations by utilizing multimodal microscopy to generate labels directly instead of manual labeling. We combine scanning electron microscopy images of 347H stainless steel as training data and electron backscatter diffraction micrographs as pixel-wise labels for grain boundary detection as a semantic segmentation task. The viability of our method is evaluated by considering a set of deep CNN architectures. We demonstrate that despite producing instrumentation drift during data collection between two modes of microscopy, this method performs comparably to similar segmentation tasks that used manual labeling. Additionally, we find that naïve pixel-wise segmentation results in small gaps and missing boundaries in the predicted grain boundary map. By incorporating topological information during model training, the connectivity of the grain boundary network and segmentation performance is improved. Finally, our approach is validated by accurate computation on downstream tasks of predicting the underlying grain morphology distributions which are the ultimate quantities of interest for microstructural characterization.

36 MATERIALS SCIENCE↗

Bragg Spot Finder (BSF): a new machine-learning-aided approach to deal with spot finding for rapidly filtering diffraction pattern images

Macromolecular crystallography contributes significantly to understanding diseases and, more importantly, how to treat them by providing atomic resolution 3D structures of proteins. This is achieved by collecting X-ray diffraction images of protein crystals from important biological pathways. Spotfinders are used to detect the presence of crystals with usable data, and the spots from such crystals are the primary data used to solve the relevant structures. Having fast and accurate spot finding is essential, but recent advances in synchrotron beamlines used to generate X-ray diffraction images have brought us to the limits of what the best existing spotfinders can do. This bottleneck must be removed so spotfinder software can keep pace with the X-ray beamline hardware improvements and be able to see the weak or diffuse spots required to solve the most challenging problems encountered when working with diffraction images. In this paper, we first present Bragg Spot Detection (BSD), a large benchmark Bragg spot image dataset that contains 304 images with more than 66 000 spots. We then discuss the open source extensible U-Net-based spotfinder Bragg Spot Finder (BSF), with image pre-processing, a U-Net segmentation backbone, and post-processing that includes artifact removal and watershed segmentation. Finally, we perform experiments on the BSD benchmark and obtain results that are (in terms of accuracy) comparable to or better than those obtained with two popular spotfinder software packages ( Dozor and DIALS ), demonstrating that this is an appropriate framework to support future extensions and improvements.

36 MATERIALS SCIENCE↗

Extreme Fast Charging of Lithium-Ion Battery: Understanding Bottlenecks and Safety Issues

Currently, graphite-based lithium-ion battery (LIB) packs in the electric vehicle (EV) domain can accommodate a charging rate up to 2C, which translates into a 0.5 to 1h recharge time [1-2]. Reducing the recharge time down to 10-15 min range would reduce the gap between the internal combustion engine’s refueling time and EV’s recharge time. Minimizing this gap is seen as attractive to consumers and would eventually enhance the adoption of EVs [1-3]. Significant research activity has been going on globally to understand the scientific limitations associated with high rate charging [4-6]. Extreme fast charging (XFC) of Li-ion batteries can create LIB life and safety issues [1, 7]. Among the issues are shortened battery life and enhanced safety concerns due to potential short creation by Li dendrites [1-2]. The detection and monitoring of Li plating onset and evolution over aging is a significant challenge. In-operando detection schemes to understand the dynamics of Li plating and the role that aging has on Li plating are vital to enable fast charging of specific energy cells. A key to the effort would be understanding the interplay between materials, electrode structure, and use conditions. This presentation will discuss the key bottlenecks of aggressive fast charging (~10 min). In particular, how extreme fast charging impacts individual battery components as well as the overall battery’s performance, life, and safety will be presented in light of experimental results and analysis. The effectiveness and challenges of in-operando electrochemical detection methods in aggressive charging conditions shall also be discussed. References [1] S. Ahmed, I. Bloom, A. N. Jansen, T. R. Tanim, E. J. Dufek, A. Pesaran, A. Burnham, R. B. Carlson, F. Dias, K. Hardy,M. Keyser, C. Kreuzer, A. Markel, A. Meintz, C. Michelbacher, M. Mohanpurkar, P. A. Nelson, D. C. Robertson, D. Scoffield,M. Shirk, T. Stephens, R. Vijayagopal, J. Zhang, Enabling fast charging–A battery technology gap assessment, J Power Sources,367 (2017), 250-262 [2] T.R. Tanim, M. G. Shirk, R. L. Bewley, E. J. Dufek, B. Y. Liaw, Fast charge implications: Pack and cell analysis and comparison,, J Power Sources, 381 (2018) 56-65 [3] N. Lutsey, S. Searle, S. Chambliss, and A. Bandivadekar, Assessment of leading electric vehicle promotion activities inUnited States cities, International Council on Clean Transportation, July 2015 [4] C. Michelbacher, S. Ahmed, I. Bloom, A. Burnham, B. Carlson, F. Dias, E. J. Dufek, A. N Jansen, M. Keyser, A. Markel,A. Meintz, M. Mohanpurkar, A. Pesaran, D. Scoffield, M. Shirk, T. Stephens, T. R. Tanim, R. Vijayagopal, J. Zhang, Enablingfast charging–Introduction and overview, J Power Sources, [5] E-Mobility: New Possibilities with 800-volt charging, Porsche newsroom, https://newsroom.porsche.com/en/technology/porsche-engineering-e-power-electromobility-800-volt-charging-12720.html(accessed Sep 2017) [6] Charging interface initiative e.V. - Industry statement on future charging infrastructure, available at: http://www.charinev.org/fileadmin/Downloads/Papers_and_Regulations/CharIN_industry_statement.pdf(accessed June 2017) [7] T. Waldmann, B. Hogg,M. Kasper, S. Grolleau, C. Couceiro, K. Trad, B. P. Matadi and M. Wohlfahrt-Mehrens, Interplay ofOperational Parameters on Lithium Deposition in Lithium-Ion Cells: Systematic Measurements with Reconstructed 3-ElectrodePouch Full Cells, J. Electrochem. Soc. 163 (7) (2016) A1232-A1238

25 ENERGY STORAGE↗

Autoencoder-Based Sensor Drift Detection and Mitigation for Resilient Charging Systems

This work presents an autoencoder-based approach for sensor signal reconstruction and drift detection for charging systems. The proposed strategy is implemented within a Simulink-based system framework and evaluated under multiple operating conditions. An autoencoder with 8 neurons in the bottleneck layer is adopted, achieving accurate reconstruction across 10 variables and strong agreement with the physical sensor readings under normal conditions. In the case of a sensor fault, the autoencoder reconstruction remains closer to the expected true value compared to the corrupted measurement. Furthermore, feeding the autoencoder-reconstructed signal value back into the control framework in place of the faulty sensor signal leads to improved power monitoring. These results highlight the potential of autoencoder-based virtual sensing to extend the concept of resiliency to all components of the charging system, including sensors.

Rezende Da Costa Reis Kimpara, Renata [ORNL] (ORCI↗

Distributed Many-to-Many Protein Sequence Alignment using Sparse Matrices

Identifying similar protein sequences is a core step in many computational biology pipelines such as detection of homologous protein sequences, generation of similarity protein graphs for downstream analysis, functional annotation, and gene location. Performance and scalability of protein similarity search have proven to be a bottleneck in many bioinformatics pipelines due to increase in cheap and abundant sequencing data. This work presents a new distributed-memory software PASTIS. PASTIS relies on sparse matrix computations for efficient identification of possibly similar proteins. We use distributed sparse matrices for scalability and show that the sparse matrix infrastructure is a great fit for protein similarity search when coupled with a fully-distributed dictionary of sequences that allow remote sequence requests to be fulfilled. Our algorithm incorporates the unique bias in amino acid sequence substitution in search without altering basic sparse matrix model, and in turn, achieves ideal scaling up to millions of protein sequences.

97 MATHEMATICS AND COMPUTING↗

NASA Electronic Library System (NELS) optimization

This is a compilation of NELS (NASA Electronic Library System) Optimization progress/problem, interim, and final reports for all phases. The NELS database was examined, particularly in the memory, disk contention, and CPU, to discover bottlenecks. Methods to increase the speed of NELS code were investigated. The tasks included restructuring the existing code to interact with others more effectively. An error reporting code to help detect and remove bugs in the NELS was added. Report writing tools were recommended to integrate with the ASV3 system. The Oracle database management system and tools were to be installed on a Sun workstation, intended for demonstration purposes.

Pribyl, William L.↗

Development of a high-throughput method for processing sponge-stick samples to detect viable Bacillus anthracis spores

Since the national validation of the sponge-stick based method for detection of Bacillus anthracis spores in environmental samples, there have not been focused efforts to address the low throughput nature of the method, which processes only one sample at one time. Sample processing remains a serious bottleneck for rapidly analyzing large numbers of samples expected from a biological warfare attack. Therefore, we developed a high-throughput method to simultaneously process multiple sponge-stick samples to be better prepared for rapid response and recovery after wide area anthrax incidents. In this method, sponges are placed in 50 mL tubes containing 25 mL extraction buffer and shaken to release spores, after which the suspension is recovered for analysis. Here, we determined that an additional extraction step, conducted in the same tubes with 10 mL buffer, further increased spore recovery from sponge-stick by approximately 10 %. We determined that orbital shaking and multi-tube vortexing were both more effective than reciprocating shaking for recovering spores. We conducted simultaneous processing of up to 12 sponge-stick samples and demonstrated comparable spore recovery efficiencies to the traditional low-throughput stomacher-based method (approximately 60 % recovery at 10 2 -spore level and 75 % recovery at 10 4 -spore level for both methods in three replicate experiments, P > 0.05 for two-tailed t-tests for each experiment and spore level). We also demonstrated that our high-throughput method could be integrated with Rapid Viability-Polymerase Chain Reaction (RV-PCR) analysis and could detect levels as low as 40 spores per sponge even when challenged by a PCR particulate contaminant.

Anthrax↗

Trigger Detection for Adaptive Scientific Workflows Using Percentile Sampling

The increasing complexity of both scientific simulations and high-performance computing system architectures are driving the need for adaptive workflows, in which the composition and execution of computational and data manipulation steps dynamically depend on the evolutionary state of the simulation itself. Consider, for example, the frequency of data storage. Critical phases of the simulation should be captured with high frequency and with high fidelity for postanalysis; however, we cannot afford to retain the same frequency for the full simulation due to the high cost of data movement. We can instead look for triggers, indicators that the simulation will be entering a critical phase, and adapt the workflow accordingly. In this paper, we present a methodology for detecting triggers and demonstrate its use in the context of direct numerical simulations of turbulent combustion using S3D. We show that chemical explosive mode analysis (CEMA) can be used to devise a noise-tolerant indicator for rapid increase in heat release. However, exhaustive computation of CEMA values dominates the total simulation, and thus is prohibitively expensive. To overcome this computational bottleneck, we propose a quantile sampling approach. Our sampling-based algorithm comes with provable error/confidence bounds, as a function of the number of samples. Most importantly, the number of samples is independent of the problem size, and thus our proposed sampling algorithm offers perfect scalability. Our experiments on homogeneous charge compression ignition and reactivity controlled compression ignition simulations show that the proposed method can detect rapid increases in heat release, and its computational overhead is negligible. Our results will be used to make dynamic workflow decisions regarding data storage and mesh resolution in future combustion simulations.

97 MATHEMATICS AND COMPUTING↗

Systems-Level Analysis of Oxygen Exposure in Zymomonas mobilis : Implications for Isoprenoid Production

ABSTRACT Zymomonas mobilis is an aerotolerant anaerobe and prolific ethanologen with attractive characteristics for industrial bioproduct generation. However, there is currently insufficient knowledge of the impact that environmental factors have on flux through industrially relevant biosynthetic pathways. Here, we examined the effect of oxygen exposure on metabolism and gene expression in Z. mobilis by combining targeted metabolomics, mRNA sequencing, and shotgun proteomics. We found that exposure to oxygen profoundly influenced metabolism, inducing both transient metabolic bottlenecks and long-term metabolic remodeling. In particular, oxygen induced a severe but temporary metabolic bottleneck in the methyl erythritol 4-phosphate pathway for isoprenoid biosynthesis caused by oxidative damage to the iron-sulfur cofactors of the final two enzymes in the pathway. This bottleneck was resolved with minimal changes in expression of isoprenoid biosynthetic enzymes. Instead, it was associated with pronounced upregulation of enzymes related to iron-sulfur cluster maintenance and biogenesis (i.e., flavodoxin reductase and the suf operon). We also detected major changes in glucose utilization in the presence of oxygen. Specifically, we observed increased gluconate production following exposure to oxygen, accounting for 18% of glucose uptake. Our results suggest that under aerobic conditions, electrons derived from the oxidation of glucose to gluconate are diverted to the electron transport chain, where they can minimize oxidative damage by reducing reactive oxygen species such as H 2 O 2 . This model is supported by the simultaneous upregulation of three membrane-bound dehydrogenases, cytochrome c peroxidase, and a cytochrome bd oxidase following exposure to oxygen. IMPORTANCE Microbially generated biofuels and bioproducts have the potential to provide a more environmentally sustainable alternative to fossil-fuel-derived products. In particular, isoprenoids, a diverse class of natural products, are chemically suitable for use as high-grade transport fuels and other commodity molecules. However, metabolic engineering for increased production of isoprenoids and other bioproducts is limited by an incomplete understanding of factors that control flux through biosynthetic pathways. Here, we examined the native regulation of the isoprenoid biosynthetic pathway in the biofuel producer Zymomonas mobilis . We leveraged oxygen exposure as a means to perturb carbon flux, allowing us to observe the formation and resolution of a metabolic bottleneck in the pathway. Our multi-omics analysis of this perturbation enabled us to identify key auxiliary enzymes whose expression correlates with increased production of isoprenoid precursors, which we propose as potential targets for future metabolic engineering.

59 BASIC BIOLOGICAL SCIENCES↗

Influence of the broadly neutralizing antibody VRC01 on HIV breakthrough virus populations in antibody-mediated prevention trials

In the antibody mediated prevention (AMP) trials, the broadly neutralizing antibody (bNAb) VRC01 demonstrated protective efficacy against susceptible HIV strains. To understand how VRC01 shaped breakthrough infections, deep sequencing was performed on 172 participants (>100,000 gag-Δpol and rev-env-Δnef sequences), at diagnosis and over time, in the placebo and treatment arms of the African (HVTN703/HPTN081; NCT02568215) and Americas/Europe (HVTN704/HPTN085; NCT02716675) cohorts. A high frequency of multilineage infections was detected (38%), including co-infection with both VRC01 sensitive and resistant viruses. This high frequency is largely accounted for by low-abundance lineages. Although VRC01 does not significantly affect the genetic transmission bottleneck compared to placebo, higher VRC01 doses trend towards greater VRC01 neutralization differences among co-infecting lineages. Two-thirds of multilineage infections showed evidence of recombination at the diagnostic timepoint. In the treatment group there is evidence of recombinant viruses preferentially inheriting resistance-associated mutations. This study provides critical insights into viral genetic and antigenic diversity that needs to be targeted to achieve protection, and highlights the role of recombination in facilitating escape.

59 BASIC BIOLOGICAL SCIENCES↗

Neural Network-Based Sensor Validation for Turboshaft Engines

Sensor failure detection, isolation, and accommodation using a neural network approach is described. An auto-associative neural network is configured to perform dimensionality reduction on the sensor measurement vector and provide estimated sensor values. The sensor validation scheme is applied in a simulation of the T700 turboshaft engine in closed loop operation. Performance is evaluated based on the ability to detect faults correctly and maintain stable and responsive engine operation. The set of sensor outputs used for engine control forms the network input vector. Analytical redundancy is verified by training networks of successively smaller bottleneck layer sizes. Training data generation and strategy are discussed. The engine maintained stable behavior in the presence of sensor hard failures. With proper selection of fault determination thresholds, stability was maintained in the presence of sensor soft failures.

Moller, James C.↗

In Situ Probing Potassium-ion Intercalation-induced Amorphization in Crystalline Iron Phosphate Cathode Materials

Na-ion and K-ion batteries are promising alternatives for large-scale energy storage applications due to their abundance and low cost. Intercalation of these large ions could cause irreversible structural deformation and partial to complete amorphization in the crystalline electrodes. The designing of new amorphous electrodes is another route to develop electrodes to store these ions reversibly. Lack of understanding of dynamic changes in the amorphous nanostructures during battery operation is the bottleneck for further developments. Here, we report the utilization of in situ digital image correlation and in-operando X-ray diffraction (XRD) techniques to probe dynamic changes in the amorphous phase of iron phosphate during potassium intercalation. In-operando XRD demonstrates amorphization in the electrode’s nanostructure during the first charge/discharge cycle. Additionally, the ex-situ high-resolution transmission electron microscopy further confirms the amorphization after potassium insertion. In situ strain analysis detects the reversible deformation associated with redox reactions in the amorphous phases. Our approach offers new insights on the mechanisms of ion intercalation in the amorphous nanostructures which are highly potent for development of next-generation batteries.

Ozdogru, Bertan↗

Probing complex geophysical geometries with chattering dust

Abstract The modern energy economy and environmental infrastructure rely on the flow of fluids through fractures in rock. Yet this flow cannot be imaged directly because rocks are opaque to most probes. Here we apply chattering dust, or chemically reactive grains of sucrose containing pockets of pressurized carbon dioxide, to study rock fractures. As a dust grain dissolves, the pockets burst and emit acoustic signals that are detected by distributed sets of external ultrasonic sensors that track the dust movement through fracture systems. The dust particles travel through locally varying fracture apertures with varying speeds and provide information about internal fracture geometry, flow paths and bottlenecks. Chattering dust particles have an advantage over chemical sensors because they do not need to be collected, and over passive tracers because the chattering dust delineates the transport path. The current laboratory work has potential to scale up to near-borehole applications in the field.

58 GEOSCIENCES↗

Electrochemical Reduction of a Spinel-Type Manganese Oxide Cathode in Aqueous Electrolytes with Ca 2+ or Zn 2+

In this report, the feasibility of reversible Ca 2+ or Zn 2+ intercalation into a crystalline cubic spinel Mn 2 O 4 cathode has been investigated using electrochemical methods in an aqueous electrolyte. A combination of synchrotron XRD and XANES studies identified the partial structural transformation from a cubic to a tetragonally distorted spinel Mn 3 O 4 , accompanied by the reduction of Mn 4+ to Mn 3+ and Mn 2+ during discharge. TEM/EDX measurements confirmed that practically no Ca 2+ was inserted upon discharge. However, non-negligible amounts of Zn were detected after Mn 2 O 4 was reduced in the Zn 2+ electrolyte, but through the formation of secondary phases that, in some cases, appeared adjacent to the surface of a cathode particle. Furthermore this report aims to identify bottlenecks in the application of manganese oxide cathodes paired with Ca or Zn metal anodes and to justify future efforts in designing prototype multivalent batteries.

25 ENERGY STORAGE↗

Scalable Incremental Checkpointing using GPU-Accelerated De-Duplication

Writing large amounts of data concurrently to stable storage is a typical I/O pattern of many HPC workflows. This pattern introduces high I/O overheads and results in increased storage space utilization especially for workflows that need to capture the evolution of data structures with high frequency as checkpoints. In this context, many applications, such as graph pattern matching, perform sparse updates to large data structures between checkpoints. For these applications, incremental checkpointing techniques that save only the differences from one checkpoint to another can dramatically reduce the checkpoint sizes, I/O bottlenecks, and storage space utilization. However, such techniques are not without challenges: it is non-trivial to transparently determine what data has changed since a previous checkpoint and assemble the differences in a compact fashion that does not result in excessive metadata. State-of-art data reduction techniques (e.g., compression and de-duplication) have significant limitations when applied to modern HPC applications that leverage GPUs: slow at detecting the differences, generate a large amount of metadata to keep track of the differences, and ignore crucial spatiotemporal checkpoint data redundancy. This paper addresses these challenges by proposing a Merkle tree-based incremental checkpointing method to exploit GPUs' high memory bandwidth and massive parallelism. Experimental results at scale show a significant reduction of the I/O overhead and space utilization of checkpointing compared with state-of-the-art incremental checkpointing and compression techniques.

Tan, Nigel↗