Search NASA⌕ Search

SEARCH · Search NASA

Results for “Data parallel computing clusters”

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 181 records · Page 10
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Ad Hoc File Systems for High-Performance Computing

Storage backends of parallel compute clusters are still based mostly on magnetic disks, while newer and faster storage technologies such as flash-based SSDs or non-volatile random access memory (NVRAM) are deployed within compute nodes. Including these new storage technologies into scientific workflows is unfortunately today a mostly manual task, and most scientists therefore do not take advantage of the faster storage media. One approach to systematically include nodelocal SSDs or NVRAMs into scientific workflows is to deploy ad hoc file systems over a set of compute nodes, which serve as temporary storage systems for single applications or longer-running campaigns. This paper presents results from the Dagstuhl Seminar 17202 “Challenges and Opportunities of User-Level File Systems for HPC” and discusses application scenarios as well as design strategies for ad hoc file systems using node-local storage media. The discussion includes open research questions, such as how to couple ad hoc file systems with the batch scheduling environment and how to schedule stage-in and stage-out processes of data between the storage backend and the ad hoc file systems. Also presented are strategies to build ad hoc file systems by using reusable components for networking and how to improve storage device compatibility. Various interfaces and semantics are presented, for example those used by the three ad hoc file systems BeeOND, GekkoFS, and BurstFS. Their presentation covers a range from file systems running in production to cutting-edge research focusing on reaching the performance limits of the underlying devices.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2013

ENHANCED MAGNETIC COMPRESSIBILITY AND ISOTROPIC SCALE INVARIANCE AT SUB-ION LARMOR SCALES IN SOLAR WIND TURBULENCE

The anisotropic nature of solar wind magnetic turbulence fluctuations is investigated scale by scale using high cadence in situ magnetic field measurements from the Cluster and ACE spacecraft missions. The data span five decades in scales from the inertial range to the electron Larmor radius. In contrast to the inertial range, there is a successive increase toward isotropy between parallel and transverse power at scales below the ion Larmor radius, with isotropy being achieved at the electron Larmor radius. In the context of wave-mediated theories of turbulence, we show that this enhancement in magnetic fluctuations parallel to the local mean background field is qualitatively consistent with the magnetic compressibility signature of kinetic Alfven wave solutions of the linearized Vlasov equation. More generally, we discuss how these results may arise naturally due to the prominent role of the Hall term at sub-ion Larmor scales. Furthermore, computing higher-order statistics, we show that the full statistical signature of the fluctuations at scales below the ion Larmor radius is that of a single isotropic globally scale-invariant process distinct from the anisotropic statistics of the inertial range.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2011

Cheetah: A Framework for Scalable Hierarchical Collective Operations

Collective communication operations, used by many scientific applications, tend to limit overall parallel application performance and scalability. Computer systems are becoming more heterogeneous with increasing node and core-per-node counts. Also, a growing number of data-access mechanisms, of varying characteristics, are supported within a single computer system. We describe a new hierarchical collective communication framework that takes advantage of hardware-specific data-access mechanisms. It is flexible, with run-time hierarchy specification, and sharing of collective communication primitives between collective algorithms. Data buffers are shared between levels in the hierarchy reducing collective communication management overhead. We have implemented several versions of the Message Passing Interface (MPI) collective operations, MPI Barrier() and MPI Bcast(), and run experiments using up to 49, 152 processes on a Cray XT5, and a small InfiniBand based cluster. At 49, 152 processes our barrier implementation outperforms the optimized native implementation by 75%. 32 Byte and one Mega-Byte broadcasts outperform it by 62% and 11%, respectively, with better scalability characteristics. Improvements relative to the default Open MPI implementation are much larger.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2014

A scalable approach to solving dense linear algebra problems on hybrid CPU-GPU systems

Aiming to fully exploit the computing power of all CPUs and all graphics processing units (GPUs) on hybrid CPU-GPU systems to solve dense linear algebra problems, in this paper we design a class of heterogeneous tile algorithms to maximize the degree of parallelism, to minimize the communication volume, and to accommodate the heterogeneity between CPUs and GPUs. The new heterogeneous tile algorithms are executed upon our decentralized dynamic scheduling runtime system, which schedules a task graph dynamically and transfers data between compute nodes automatically. The runtime system uses a new distributed task assignment protocol to solve data dependencies between tasks without any coordination between processing units. By overlapping computation and communication through dynamic scheduling, we are able to attain scalable performance for the double-precision Cholesky factorization and QR factorization. Finally, our approach demonstrates a performance comparable to Intel MKL on shared-memory multicore systems and better performance than both vendor (e.g., Intel MKL) and open source libraries (e.g., StarPU) in the following three environments: heterogeneous clusters with GPUs, conventional clusters without GPUs, and shared-memory systems with multiple GPUs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

GPU acceleration of Swendsen–Wang dynamics

When simulating a lattice system near its critical temperature, local algorithms for modeling the system’s evolution can introduce very large autocorrelation times into sampled data. Here, this critical slowing down places restrictions on the analysis that can be completed in a timely manner of the behavior of systems around the critical point. Because it is often desirable to study such systems around this point, a new algorithm must be introduced. Therefore, we turn to cluster algorithms, such as the Swendsen–Wang algorithm and the Wolff clustering algorithm. They incorporate global updates which generate new lattice configurations with little correlation to previous states, even near the critical point. We look to accelerate the rate at which these algorithm are capable of running by implementing and benchmarking a parallel implementation of each algorithm designed to run on GPUs under NVIDIA’s CUDA framework. A 17 and 90 fold increase in the computational rate was, respectively, experienced when measured against the equivalent algorithm implemented in serial code.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2019

A cloud computing approach to massively-parallel distributed stochastic optimization with application to geomechanical reservoir characterization

Many search strategies used to optimize parameters during hydrologic model calibration are inherently limited in that simulations must be performed sequentially in order for the algorithm to converge. As a result, the overall time required to reach a solution increases linearly with the number of steps necessary. In contrast, sampling strategies that attempt to approximate the posterior distribution of the parameters do not require the model simulations to be run in sequence and therefore invite the use of a massively parallel approach across heterogeneous computing environments. We apply a cloud computing methodology to this problem where a head node running on Amazon Web Services continuously produces candidate models that are distributed to arbitrary compute environments to run the simulation, in our case Clemson’s Palmetto cluster computer. Thus the time required to obtain a sufficient number of samples to produce an adequate approximation of the posterior distribution of the parameters is limited by the number of available compute sites. Monte Carlo sampling of the posterior space would be duplicative and inefficient, thus we combine Latin hypercube methods with genetic algorithms to ensure that no large regions of the parameter space are left unexplored, and that high probability regions are investigated thoroughly. We independently assess the fit to different types of data sets without the need to construct a single objective function. Sets of Pareto optimal models are particularly valuable when the quality and information content of data is poorly known or conceptual model errors exist in our simulation. We illustrate our approach for a problem where in situ strain data are used to calibrate the parameters of an oil reservoir in the Avant oil field in Oklahoma. We use well injection and shut-in to stress a deep (500m) confined formation, and observe the geomechanical response using borehole strainmeters installed in the shallow subsurface (30m). These data were then used to calibrate the poroelastic properties of the reservoir and caprock along with the geometry of an elliptic channel associated with the formation. Using our approach we are able to identify issues with model error as well as substantial tradeoffs that occur within the parameters.

inversion, stochastic, geomechanics, poroelastic, ↗
DOE OSTI2012

Data-Parallel Halo Finder Operator in PISTON

PISTON is a portable framework which supports the development of visualization and analysis operators using a platform-independent, data-parallel programming model. Operators such as isosurface, cut-surface and threshold have been implemented in this framework, with the exact same operator code achieving good parallel performance on different architectures. An important analysis operator in cosmology is the halo finder. A halo is a cluster of particles and is considered a common feature of interest found in cosmology data. As the number of cosmological simulations carried out in the recent past has increased, the resultant data of these simulations and the required analysis tasks have increased as well. As a consequence, there is a need to develop scalable and efficient tools to carry out the needed analysis. Therefore, we are currently implementing a halo finder operator using PISTON. Researchers have developed a wide variety of techniques to identify halos in raw particle data. The most basic algorithm is the friend-of-friends (FOF) halo finder, where the particles are clustered based on two parameters: linking length and halo size. In a FOF halo finder, all particles which lie within the linking length are considered as one halo and the halos are filtered based on the halo size parameter. A naive implementation of a FOF halo finder compares each and every particle pair, requiring O(n{sup 2}) operations. Our data-parallel halo finder operator uses a balanced k-d tree to reduce this number of operations in the average case, and implements the algorithm using only the data-parallel primitives in order to achieve portability and performance.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2017

Data Structures for Extreme Scale Computing

As computing problems of national importance grow, the government meets the increased demand by funding the development of ever larger systems. The overarching goal of the work supported in part by this grant is to increase efficiency of programming and performing computations on these large computing systems. In past work, we have demonstrated that some of these computations once thought to require expensive hardware designs and/or complex, special-purpose programming may be executed efficiently on low-cost commodity cluster computing systems using a general-purpose “latency-tolerant” programming framework. One important developed application of the ideas underlying this framework is graph database technology supporting social network pattern matching used by US intelligence agencies to more quickly identify potential terrorist threats. This database application has been spun out by the Pacific Northwest National Laboratory, a Department of Energy Laboratory, into a commercial start-up, Trovares Inc. We explore an alternative application of the same underlying ideas to a well-studied challenge arising in engineering: solving unstructured sparse linear equations. Solving these equations is key to predicting the behavior of large electronic circuits before they are fabricated. Predicting that behavior ahead of fabrication means that designs can optimized and errors corrected ahead of the expense of manufacture.

42 ENGINEERING↗
DOE OSTI2020

Deep Generative Models that Solve PDEs: Distributed Computing for Training Large Data-Free Models

Recent progress in scientific machine learning (SciML) has opened up the possibility of training novel neural network architectures that solve complex partial differential equations (PDEs). Several (nearly data free) approaches have been recently reported that successfully solve PDEs, with examples including deep feed forward networks, generative networks, and deep encoder-decoder networks. However, practical adoption of these approaches is limited by the difficulty in training these models, especially to make predictions at large output resolutions (≥1024×1024). Here we report on a software framework for data parallel distributed deep learning that resolves the twin challenges of training these large SciML models - training in reasonable time as well as distributing the storage requirements. Our framework provides several out of the box functionality including (a) loss integrity independent of number of processes, (b) synchronized batch normalization, and (c) distributed higher-order optimization methods. We show excellent scalability of this framework on both cloud as well as HPC clusters, and report on the interplay between bandwidth, network topology and bare metal vs cloud. We deploy this approach to train generative models of sizes hitherto not possible, showing that neural PDE solvers can be viably trained for practical applications. We also demonstrate that distributed higher-order optimization methods are 2-3× faster than stochastic gradient-based methods and provide minimal convergence drift with higher batch-size.

PDEs↗
DOE OSTI2012

Massively parallel X-ray scattering simulations

Although present X-ray scattering techniques can provide tremendous information on the nano-structural properties of materials that are valuable in the design and fabrication of energy-relevant nano-devices, a primary challenge remains in the analyses of such data. In this paper we describe a high-performance, flexible, and scalable Grazing Incidence Small Angle X-ray Scattering simulation algorithm and codes that we have developed on multi-core/CPU and many-core/GPU clusters. We discuss in detail our implementation, optimization and performance on these platforms. Our results show speedups of ~125x on a Fermi-GPU and ~20x on a Cray-XE6 24-core node, compared to a sequential CPU code, with near linear scaling on multi-node clusters. To our knowledge, this is the first GISAXS simulation code that is flexible to compute scattered light intensities in all spatial directions allowing full reconstruction of GISAXS patterns for any complex structures and with highresolutions while reducing simulation times from months to minutes.

Sarje, Abhinav↗
DOE OSTI2011

Designing a Scalable Fault Tolerance Model for High Performance Computational Chemistry: A Case Study with Coupled Cluster Perturbative Triples

In the last couple of decades, the massive computational power provided by the most modern supercomputers has resulted in simulation of higher order computational chem- istry methods, previously considered intractable. As the system sizes continue to increase, computational chemistry domain continues to escalate this trend using parallel comput- ing with programming models such as Message Passing Interface (MPI) and Partitioned Global Address Space (PGAS) programming models such as Global Arrays. The ever increasing scale of these supercomputers comes at a cost of reduced mean time between failures, currently in the order of days, and projected to be in the order of hours for up- coming extreme scale systems. While traditional disk based checkpointing methods are ubiquitous for storing intermediate solutions, they suffer from high overhead of writing and recovering from checkpoints. In practice, checkpointing itself often brings the system down. Clearly, methods beyond checkpointing are imperative to handling the aggravating issue of reducing MTBF. In this paper, we address this challenge by designing and im- plementing an efficient fault tolerant version of coupled cluster method with NWChem, using in memory data redundancy. We present the challenges associated with our de- sign including efficient data storage model, maintenance of at least one consistent data copy and the recovery process. Our performance evaluation without faults shows that the current design exhibits negligible overhead. In the presence of a fault, the proposed design incurs negligible overhead in comparison to the state of the art implementation without faults.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2012

Applied Parallel Metadata Indexing

The GPFS Archive is parallel archive is a parallel archive used by hundreds of users in the Turquoise collaboration network. It houses 4+ petabytes of data in more than 170 million files. Currently, users must navigate the file system to retrieve their data, requiring them to remember file paths and names. A better solution might allow users to tag data with meaningful labels and searach the archive using standard and user-defined metadata, while maintaining security. last summer, I developed the backend to a tool that adheres to these design goals. The backend works by importing GPFS metadata into a MongoDB cluster, which is then indexed on each attribute. This summer, the author implemented security and developed the user interfae for the search tool. To meet security requirements, each database table is associated with a single user, which only stores records that the user may read, and requires a set of credentials to access. The interface to the search tool is implemented using FUSE (Filesystem in USErspace). FUSE is an intermediate layer that intercepts file system calls and allows the developer to redefine how those calls behave. In the case of this tool, FUSE interfaces with MongoDB to issue queries and populate output. A FUSE implementation is desirable because it allows users to interact with the search tool using commands they are already familiar with. These security and interface additions are essential for a usable product.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2011

Provisioning a Multi-Tiered Data Staging Area for Extreme-Scale Machines

Massively parallel scientific applications, running on extreme-scale supercomputers, produce hundreds of terabytes of data per run, driving the need for storage solutions to improve their I/O performance. Traditional parallel file systems (PFS) in high performance computing (HPC) systems are unable to keep up with such high data rates, creating a storage wall. In this work, we present a novel multi-tiered storage architecture comprising hybrid node-local resources to construct a dynamic data staging area for extreme-scale machines. Such a staging ground serves as an impedance matching device between applications and the PFS. Our solution combines diverse resources (e.g., DRAM, SSD) in such a way as to approach the performance of the fastest component technology and the cost of the least expensive one. We have developed an automated provisioning algorithm that aids in meeting the checkpointing performance requirement of HPC applications, by using a least-cost storage configuration. We evaluate our approach using both an implementation on a large scale cluster and a simulation driven by six-years worth of Jaguar supercomputer job-logs, and show that our approach, by choosing an appropriate storage configuration, achieves 41.5% cost savings with only negligible impact on performance.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2017

Exploring DataVortex Systems for Irregular Applications

Emerging applications for data analytics and knowledge discovery typically have irregular or unpredictable communication patterns that do not scale well on parallel systems designed for traditional bulk-synchronous HPC applications. New network architectures that focus on minimizing (short) message latencies, rather than maximizing (large) transfer bandwidths, are emerging as possible alternatives to better support those applications with irregular communication patterns. We explore a system based upon one such novel network architecture, the Data Vortex interconnection network, and examine how this system performs by running benchmark code written for the Data Vortex network, as well as a reference MPI-over- Infiniband implementation, on the same cluster.

Data Vortex, High-Performance Computing, Networkin↗
DOE OSTI2017

Preparing NERSC users for Cori, a Cray XC40 system with Intel many integrated cores

The newest NERSC supercomputer Cori is a Cray XC40 system consisting of 2,388 Intel Xeon Haswell nodes and 9,688 Intel Xeon-Phi “Knights Landing” (KNL) nodes. Compared to the Xeon-based clusters NERSC users are familiar with, optimal performance on Cori requires consideration of KNL mode settings; process, thread, and memory affinity; fine-grain parallelization; vectorization; and use of the high-bandwidth MCDRAM memory. This paper describes our efforts preparing NERSC users for KNL through the NERSC Exascale Science Application Program, Web documentation, and user training. We discuss how we configured the Cori system for usability and productivity, addressing programming concerns, batch system configurations, and default KNL cluster and memory modes. Here, system usage data, job completion analysis, programming and running jobs issues, and a few successful user stories on KNL are presented.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Computing the Properties of Matter with Leadership Computing Resources (Closeout Report for DE-SC0018121)

In order to add more capabilities to Halide, we have designed a new framework called Tiramisu and integrated this framework into Halide. Since Tiramisu enables Halide to target heterogeneous architectures, our development efforts have been refocused on Tiramisu. Most high-performance computer systems today are complex and increasingly heterogeneous; they may have CPUs, GPUs and FPGAs. Achieving best performance requires taking full advantage of all these different architectures. To address this issue, we have designed Tiramisu, an optimization framework that enables Halide (and other DSLs) to target heterogeneous architectures. Tiramisu is an optimization framework that takes as input a high level, architecture-independent representation of code and a set of scheduling and data mapping commands that guide code transformation. The input can either be generated by a domain-specific language (DSL) compiler such as Halide or directly written by a programmer. Tiramisu then applies the user-specified code and data-layout transformations and generates an architecture-specific, low-level intermediate representation (IR) that takes advantage of modern architectural features such as multicore parallelism, non-uniform memory (NUMA) hierarchies, clusters, and accelerators like GPUs and FPGAs. We integrated Tiramisu within Halide and implemented a representative set of benchmarks to evaluate this integration. Tiramisu is now open source and is available for public use (http://tiramisu-compiler.org/). A paper about Tiramisu was published, it shows that Tiramisu extends Halide with many new capabilities and that Tiramisu can generate efficient code for multicores, GPUs, FPGAs and distributed heterogeneous systems. The performance of code generated by the Tiramisu backends matches or exceeds hand optimized reference implementations. For example, the multicore backend matches the highly optimized Intel MKL library on many kernels and shows speedups reaching 4x over the original Halide. In addition to making Tiramisu more robust, we have used Tiramisu to implement a set of representative tensor operation for constructing baryon building blocks required for multi baryon contractions in LQCD. In order to implement this code, we needed to generalize Tiramisu in two ways: first we needed to support indirect array accesses, and second, we needed to add support for complex numbers to Tiramisu. The code generated by Tiramisu is 6x faster than the reference code. Our efforts towards an MPI based multi-node version of tiramisu have matured and the resulting code scales well on multiple nodes (tests up to 512 KNL nodes have been undertaken).

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

ROOT RNTuple and EOS: The Next Generation of Event Data I/O

For several years, the ROOT team is developing the new RNTuple I/O subsystem in preparation of the next generation of collider experiments. Both HL-LHC and DUNE are expected to start data taking by the end of this decade. They pose unprecedented challenges to event data I/O in terms of data rates, event sizes, and event complexity. At the same time, the I/O landscape is becoming more diverse. HPC cluster file systems and object stores, NVMe disk cache layers in analysis facilities, and S3 storage on cloud resources are mixing with traditional XRootD-managed spinning disk pools.The ROOT team will finalize a first production version of the RNTuple binary format by the end of 2024. After this point, ROOT will provide backward compatibility for RNTuple data. This contribution provides an overview of the RNTuple feature set, the related R&D activities and the long-term vision for RNTuple. We report on performance, interface design, tooling, robustness, integration with experiment frameworks, and validation results, as well as recent R&D on parallel reading and writing and exploitation of modern hardware and storage systems. We will give an outlook on possible future features after a first production release.Collaboratively, the IT and EP departments at CERN have launched a formal project within the Research and Computing sector to evaluate the novel data format for physics analysis data utilized in LHC experiments and other fields. This part of the project focuses on validating the scalability of the EOS storage backend during the transition from the over 25 years old TTree production format to the newly developed RNTuple format, using both replicated and erasure-coded storage profiles.

Blomer, Jakob [CERN]↗