Search NASA⌕ Search

SEARCH · Search NASA

Results for “parallel clustering”

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 397 records · Page 22

Scalable Distributed Memory Community Detection Using Vite

Graph clustering, popularly known as community detection, is a fundamental graph operation used in several applications of relevance for network analysis and cybersecurity. Community detection is an operation of partitioning a network into an arbitrary number of “communities” such that each com munity represents a tightly-knit group of nodes with relatively sparser connections to the rest of the nodes in the network. The need to compute clustering on large-scale networks necessitates the development of efficient parallel algorithms capable of exploiting modern parallel architectures. However, due to their irregular and inherently sequential nature, many of the current algorithms for community detection are challenging to parallelize. In response to the 2018 Graph Challenge, we present Vite which is a distributed-memory parallel community detection method. Vite is a parallelization of the algorithmic template provided by a widely-used serial method (Louvain). Alongside, it uses several heuristics aimed at significantly improving parallelism and performance, while preserving output quality. Using the inputs from the 2018 Graph Challenge (static and streaming), we demonstrate superior performance and high quality solutions based on four parallelization heuristics.

community detection↗

DAAP (Data Analytics Application Profiling)

The Data Analytics Application Profiling (DAAP) utility is a software program comprised of a library and scripts to provide basic instrumenting of parallel scientific software applications while they are running on an HPC cluster, and transmit this data over a secure connection as individual records to aggregators. When used in conjunction with additional open source software components (Telegraf, FluentD, and RabbitMQ), the data collected by DAAP can be transmitted off-cluster to an analytics system of the user’s choice for further analysis.

Shereda, Charles↗

Exploring Flexible Communications for Streamlining DNN Ensemble Training Pipelines

Parallel training of a Deep Neural Network (DNN) ensemble on a cluster of nodes is a common practice to train multiple models in order to construct a model with a higher prediction accuracy. Existing ensemble training pipelines can perform a great deal of redundant operations, resulting in unnecessary CPU usage, or even poor pipeline performance. In order to remove these redundancies, we need pipelines with more communication flexibility than existing DNN frameworks can provide. This project investigates a series of designs to improve pipeline flexibility and adaptivity, while also increasing performance. We implement our designs using Tensorflow with Horovod, and test it using several large DNNs in a large scale GPU cluster, Titan supercomputer at Oak Ridge National Lab. Our results show that the CPU time spent during training is reduced by 2-11X. Furthermore, our implementation can achieve up to 10X speedups when CPU core limits are imposed. Our best pipeline also reduces the average power draw of the ensemble training process by 5-16% when compared to the baseline.

97 MATHEMATICS AND COMPUTING↗

Parallel architectures for iterative methods on adaptive, block structured grids

A parallel computer architecture well suited to the solution of partial differential equations in complicated geometries is proposed. Algorithms for partial differential equations contain a great deal of parallelism. But this parallelism can be difficult to exploit, particularly on complex problems. One approach to extraction of this parallelism is the use of special purpose architectures tuned to a given problem class. The architecture proposed here is tuned to boundary value problems on complex domains. An adaptive elliptic algorithm which maps effectively onto the proposed architecture is considered in detail. Two levels of parallelism are exploited by the proposed architecture. First, by making use of the freedom one has in grid generation, one can construct grids which are locally regular, permitting a one to one mapping of grids to systolic style processor arrays, at least over small regions. All local parallelism can be extracted by this approach. Second, though there may be a regular global structure to the grids constructed, there will be parallelism at this level. One approach to finding and exploiting this parallelism is to use an architecture having a number of processor clusters connected by a switching network. The use of such a network creates a highly flexible architecture which automatically configures to the problem being solved.

Gannon, D.↗

Experimental studies of the properties of 'simulated' upstream turbulence using a statistical multipoint method

In this report we present a different approach to the multipoint measurement of magnetic fields and plasma. This is called the multi-spacecraft ensemble technique (MET), essentially free of process restrictions, such as linearity and stationarity. We comprehensively discuss the other conditions and limitations intrinsic to this statistical method. We also show the results of the application of the ensemble method to the synthetic data obtained from a hybrid simulation in the region upstream of a quasi-parallel shock. The important implications of the above approach for the CLUSTER mission are discussed.

Orlowski, D. S.↗

Exploring flexible communications for streamlining DNN ensemble training pipelines

Parallel training of a Deep Neural Network (DNN) ensemble on a cluster of nodes is a common practice to train multiple models in order to construct a model with a higher prediction accuracy, or to quickly tune the parameters of a training model. Existing ensemble training pipelines perform a great deal of redundant operations, resulting in unnecessary CPU usage, or even poor pipeline performance. In order to remove these redundancies, we need pipelines with more communication flexibility than existing DNN frameworks can provide. This project investigates a series of designs to improve pipeline flexibility and adaptivity, while also increasing performance. We implement our designs using Tensorflow with Horovod, and test it using several large DNNs in a large scale GPU cluster, the Titan supercomputer at Oak Ridge National Lab. Our results show that with the new flexible communication schemes, the CPU time spent during training is reduced by 2-11X. Furthermore, our implementation can achieve up to 10X speedups when CPU core limits are imposed. Our best pipeline also reduces the average power draw of the ensemble training process by 5--16% when compared to the baseline.

Pittman, Randall↗

FLEET: Flexible Efficient Ensemble Training for Heterogeneous Deep Neural Networks

Parallel training of an ensemble of Deep Neural Networks (DNN) on a cluster of nodes is an effective approach to shorten the process of neural network architecture search and hyper-parameter tuning for a given learning task. Prior efforts have shown that data sharing, where the common preprocessing operation is shared across the DNN training pipelines, saves computational resources and improves pipeline efficiency. Data sharing strategy, however, performs poorly for a heterogeneous set of DNNs where each DNN has varying computational needs and thus different training rate and convergence speed. This paper proposes FLEET, a flexible ensemble DNN training framework for efficiently training a heterogeneous set of DNNs. We build FLEET via several technical innovations. We theoretically prove that an optimal resource allocation is NP-hard and propose a greedy algorithm to efficiently allocate resources for training each DNN with data sharing. We integrate data-parallel DNN training into ensemble training to mitigate the differences in training rates and introduce checkpointing into this context to address the issue of different convergence speeds. Experiments show that FLEET significantly improves the training efficiency of DNN ensembles without compromising the quality of the result.

Guan, Hui↗

Massively parallel quantum chemical density matrix renormalization group method

Here, we present, to the best of our knowledge, the first attempt to exploit the supercomputer platform for quantum chemical density matrix renormalization group (QCDMRG) calculations. We have developed the parallel scheme based on the in-house MPI global memory library, which combines operator and symmetry sector parallelisms, and tested its performance on three different molecules, all typical candidates for QC-DMRG calculations. In case of the largest calculation, which is the nitrogenase FeMo cofactor cluster with the active space comprising 113 electrons in 76 orbitals and bond dimension equal to 6000, our parallel approach scales up to approximately 2000 CPU cores.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An Approach to Selecting Thread + Process Mixes for Hybrid MPI + OpenMP Applications, In: 2015 IEEE International Conference on Cluster Computing

Hybrid MPI + OpenMP is a popular means of programming modern machines that feature substantial parallelism both off-node and on-node. Determining the right mix of the two programming models to use, however, is not as straightforward as simply using exclusively OpenMP on-node and limiting MPI to only inter-node communication. We present a step-by-step methodology to help make the decision of which mix of the two programming models to use. It starts with an estimate of the performance of a generic hybrid application on a given machine and incorporates additional available information about the specific application and the machine to provide guidance for selecting effective mixes of MPI processes and OpenMP threads to use when running that application on the machine in question. We validate our approach on four different applications on an IBM Blue Gene/Q, a Cray XK7, and a Cray XC30.

Computer Science↗

Observations of parallel ion energization in the equatorial region

In this paper an example of bounce-phase-clustered ions such as earlier reported by Mellwain (1976), Quinn (1978) and Quinn and Mellwain (1979) is examined in detail. The cluster has small pitch angle, but investigation of the energy time dispersion properties of the cluster reveals that it was formed at the equator. The acceleration region extended over about 2.4 earth radii along the field direction. It is suggested that the ion cluster is set up by acceleration during a transient convection surge, and the properties of such a process are discussed.

Quinn, J. M.↗

Scaling Semantic Graph Databases in Size and Performance

In this paper we present SGEM, a full software system for accelerating large-scale semantic graph databases on commodity clusters. Unlike current approaches, SGEM addresses semantic graph databases by only employing graph methods at all the levels of the stack. On one hand, this allows exploiting the space efficiency of graph data structures and the inherent parallelism of graph algorithms. These features adapt well to the increasing system memory and core counts of modern commodity clusters. On the other hand, however, these systems are optimized for regular computation and batched data transfers, while graph methods usually are irregular and generate fine-grained data accesses with poor spatial and temporal locality. Our framework comprises a SPARQL to data parallel C compiler, a library of parallel graph methods and a custom, multithreaded runtime system. We introduce our stack, motivate its advantages with respect to other solutions and show how we solved the challenges posed by irregular behaviors. We present the result of our software stack on the Berlin SPARQL benchmarks with datasets up to 10 billion triples (a triple corresponds to a graph edge), demonstrating scaling in dataset size and in performance as more nodes are added to the cluster.

Aggregation↗

Layout-aware I/O Scheduling for terabits data movement

Many science facilities, such as the Department of Energy's Leadership Computing Facilities and experimental facilities including the Spallation Neutron Source, Stanford Linear Accelerator Center, and Advanced Photon Source, produce massive amounts of experimental and simulation data. These data are often shared among the facilities and with collaborating institutions. Moving large datasets over the wide-area network (WAN) is a major problem inhibiting collaboration. Next-generation, terabit-networks will help alleviate the problem, however, the parallel storage systems on the endsystem hosts at these institutions can become a bottleneck for terabit data movement. The parallel storage system (PFS) is shared by simulation systems, experimental systems, analysis and visualization clusters, in addition to wide-area data movers. These competing uses often induce temporary, but significant, I/O load imbalances on the storage system, which impact the performance of all the users. The problem is a serious concern because some resources are more expensive (e.g. super computers) or have time-critical deadlines (e.g. experimental data from a light source), but parallel file systems handle all requests fairly even if some storage servers are under heavy load. This paper investigates the problem of competing workloads accessing the parallel file system and how the performance of wide-area data movement can be improved in these environments. First, we study the I/O load imbalance problems using actual I/O performance data collected from the Spider storage system at the Oak Ridge Leadership Computing Facility. Second, we present I/O optimization solutions with layout-awareness on end-system hosts for bulk data movement. With our evaluation, we show that our I/O optimization techniques can avoid the I/O congested disk groups, improving storage I/O times on parallel storage systems for terabit data movement.

Kim, Youngjae↗

Layout-Aware I/O Scheduling for Terabits Data Movement

Many science facilities, such as the Department of Energy s Leadership Computing Facilities and experimental facilities including the Spallation Neutron Source, Stanford Linear Accelerator Center, and Advanced Photon Source, produce massive amounts of experimental and simulation data. These data are often shared among the facilities and with collaborating institutions. Moving large datasets over the wide- area network (WAN) is a major problem inhibiting collaboration. Next- generation, terabit-networks will help alleviate the problem, however, the parallel storage systems on the end-system hosts at these institutions can become a bottleneck for terabit data movement. The parallel storage system (PFS) is shared by simulation systems, experimental systems, analysis and visualization clusters, in addition to wide-area data movers. These competing uses often induce temporary, but significant, I/O load imbalances on the storage system, which impact the performance of all the users. The problem is a serious concern because some resources are more expensive (e.g. super computers) or have time-critical deadlines (e.g. experimental data from a light source), but parallel file systems handle all requests fairly even if some storage servers are under heavy load. This paper investigates the problem of competing workloads accessing the parallel file system and how the performance of wide-area data movement can be improved in these environments. First, we study the I/O load imbalance problems using actual I/O performance data collected from the Spider storage system at the Oak Ridge Leadership Computing Facility. Second, we present I/O optimization solutions with layout-awareness on end-system hosts for bulk data movement. With our evaluation, we show that our I/O optimization techniques can avoid the I/O congested disk groups, improving storage I/O times on parallel storage systems for terabit data movement.

I/O and Storage↗

Exceptionally Complex Electronic Structures of Lanthanide Oxides and Small Molecules

Lanthanide (Ln) oxide clusters and molecular systems provide a bottom-up look at the electronic structures of the bulk materials because of close parallels in the patterns of Ln 4fN subshell occupancy between the molecular and bulk Ln 2 O 3 size limits. At the same time, these clusters and molecules offer a challenge to the theory community to find appropriate and robust treatments for the 4f N patterns across the Ln series. Anion photoelectron (PE) spectroscopy provides a powerful experimental tool for studying these systems, mapping the energies of the ground and low-lying excited states of the neutral relative to the initial anion state, providing spectroscopic patterns that reflect the Ln 4f N occupancy. In this Account, we review our anion PE spectroscopic and computational studies on a range of small lanthanide molecules and cluster species. The PE spectra of LnO – (Ln = Ce, Pr, Sm, Eu) diatomic molecules show spectroscopic signatures associated with detachment of an electron from what can be described as a diffuse Ln 6s-like orbital. While the spectra of all four diatomics share this common transition, the fine structure in the transition becomes more complex with increasing 4f occupancy. This effect reflects increased coupling between the electrons occupying the corelike 4f and diffuse 6s orbitals with increasing N. Understanding the PE spectra of these diatomics sets the stage for interpreting the spectra of polyatomic molecular and cluster species. In general, the results confirm that the partial 4f N subshell occupancy is largely preserved between molecular and bulk oxides and borides. However, they also suggest that surfaces and edges of bulk materials may support a low-energy, diffuse Ln 6s band, in contrast to bulk interiors, in which the 6s band is destabilized relative to the 5d band. We also identify cases in which the molecular Ln centers have 4f N+1 occupancy rather than bulklike 4f N , which results in weaker Ln–O bonding. Specifically, Sm centers in mixed Ce–Sm oxides or in Sm x O y – (y ≤ x) clusters have this higher 4f N+1 occupancy. The PE spectra of these particular species exhibit a striking increase in the relative intensities of excited-state transitions with decreasing photon energy (resulting in lower photoelectron kinetic energy). Furthermore, this is opposite of what is expected on the basis of the threshold laws that govern photodetachment. We relate this phenomenon to strong electron–neutral interactions unique to these complex electronic structures. The time scale of the interaction, which shakes up the electronic configuration of the neutral, increases with decreasing electron momentum. From a computational standpoint, we point out that special care must be taken when considering Ln cluster and molecular systems toward the center of the Ln series (e.g., Sm, Eu), where treatment of electrons explicitly or using an effective core potential can yield conflicting results on competing subshell occupancies. However, despite the complex electronic structures associated with partially filled 4f N subshells, we demonstrate that inexpensive and tractable calculations yield useful qualitative insight into the general electronic structural features.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Visualization of unsteady computational fluid dynamics

The current computing environment that most researchers are using for the calculation of 3D unsteady Computational Fluid Dynamic (CFD) results is a super-computer class machine. The Massively Parallel Processors (MPP's) such as the 160 node IBM SP2 at NAS and clusters of workstations acting as a single MPP (like NAS's SGI Power-Challenge array) provide the required computation bandwidth for CFD calculations of transient problems. Work is in progress on a set of software tools designed specifically to address visualizing 3D unsteady CFD results in these super-computer-like environments. The visualization is concurrently executed with the CFD solver. The parallel version of Visual3, pV3 required splitting up the unsteady visualization task to allow execution across a network of workstation(s) and compute servers. In this computing model, the network is almost always the bottleneck so much of the effort involved techniques to reduce the size of the data transferred between machines.

Haimes, Robert↗

Revisiting benzene cluster cations for the chemical ionization of dimethyl sulfide and select volatile organic compounds

Benzene cluster cations were revisited as a sensitive and selective reagent ion for the chemical ionization of dimethyl sulfide (DMS) and a select group of volatile organic compounds (VOCs). Laboratory characterization was performed using both a new set of compounds (i.e., DMS, β -caryophyllene) as well as previously studied VOCs (i.e., isoprene, α -pinene). Using a field deployable chemical-ionization time-of-flight mass spectrometer (CI-ToFMS), benzene cluster cations demonstrated high sensitivity (> 1 ncps ppt &minus;1 ) to DMS, isoprene, and α -pinene standards. Parallel measurements conducted using a chemical-ionization quadrupole mass spectrometer, with a much weaker electric field, demonstrated that ion–molecule reactions likely proceed through a combination of ligand-switching and direct charge transfer mechanisms. Laboratory tests suggest that benzene cluster cations may be suitable for the selective ionization of sesquiterpenes, where minimal fragmentation (< 25 %) was observed for the detection of β -caryophyllene, a bicyclic sesquiterpene. The in-field stability of benzene cluster cations using CI-ToFMS was examined in the marine boundary layer during the High Wind Gas Exchange Study (HiWinGS). The use of benzene cluster cation chemistry for the selective detection of DMS was validated against an atmospheric pressure ionization mass spectrometer, where measurements from the two instruments were highly correlated ( R 2 > 0.95, 10 s averages) over a wide range of sampling conditions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Properties of polycyclic aromatic hydrocarbons in the northwest photon dominated region of NGC 7023. II. Traditional PAH analysis using k-means as a visualization tool

Polycyclic aromatic hydrocarbon (PAH) emission in the Spitzer-IRS spectral map of the northwest photon dominated region (PDR) in NGC 7023 is analyzed using the 'traditional' approach in which the PAH bands and plateaus between 5.2-19.5 μm are isolated by subtracting the underlying continuum and removing H{sub 2} emission lines. The spectra are organized into seven spectroscopic bins by using k-means clustering. Each cluster corresponds to, and reveals, a morphological zone within NGC 7023. The zones self-organize parallel to the well-defined PDR front that coincides with an increase in intensity of the H{sub 2} emission lines. PAH band profiles and integrated strengths are measured, classified, and mapped. The morphological zones revealed by the k-means clustering provides deeper insight into the conditions that drive variations in band strength ratios and evolution of the PAH population that otherwise would be lost. For example, certain band-band relations are bifurcated, revealing two limiting cases; one associated with the PDR, the other with the diffuse medium. Traditionally, PAH band strength ratios are used to gain insight into the properties of the emitting PAH population, i.e., charge, size, structure, and composition. Insights inferred from this work are compared and contrasted to those from Boersma et al. (first paper in this series), where the PAH emission in NGC 7023 is decomposed exclusively using the PAH spectra and tools made available through the NASA Ames PAH IR Spectroscopic Database.

79 ASTRONOMY AND ASTROPHYSICS↗

FIRST EVIDENCE OF FULLY SPATIALLY MIXED FIRST AND SECOND GENERATIONS IN GLOBULAR CLUSTERS: THE CASE OF NGC 6362

We present the first evidence of multiple populations in the Galactic globular cluster NGC 6362. We used optical and near-UV Hubble Space Telescope and ground-based photometry, finding that both the sub-giant and red giant branches are split in two parallel sequences in all color-magnitude diagrams where the F336W filter (or U band) is used. This cluster is one of the least massive globulars (M {sub tot} ∼ 5 × 10{sup 4} M {sub ☉}) where multiple populations have been detected so far. Even more interestingly and at odds with any previous finding, we observe that the two identified populations share the same radial distribution across the cluster extension. NGC 6362 is the first system where stars from different populations are found to be completely spatially mixed. Based on N-body and hydrodynamical simulations of multiple stellar generations, we argue that to reproduce these findings NGC 6362 should have lost up to 80% of its original mass.

79 ASTRONOMY AND ASTROPHYSICS↗