Search NASA⌕ Search

SEARCH · Search NASA

Results for “Load Balancing”

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 271 records · Page 15

Towards scaling community detection on distributed-memory heterogeneous systems

Distributed multi-GPU systems pose significant challenges and opportunities for efficient execution of parallel applications. Graph algorithms are generally characterized by irregular memory accesses, low computation to communication ratios, and load balancing problems that are especially hard to address on multi-GPU systems. Graph community detection is an important problem in the emerging domain of graph analytics with numerous applications. In this paper, we present our ongoing work on distributed-memory multi-GPU implementation for graph community detection. Our work parallelizes the widely used (albeit serial) Louvain method on distributed multi-GPU platforms. Supported by an extensive set of experiments on a multi-GPU enabled supercomputer (OLCF Summit) and a single compute node (Nvidia DGX-2®), we demonstrate competitive performance to existing distributed-memory CPU-based implementation, and up to 6.5 better results than Nvidia RAPIDS® CUGRAPH. To the best of our knowledge, this work represents the first effort for community detection on distributed multi-GPU systems. Our approach and related findings can be extended to numerous other iterative graph algorithms on multi-GPU systems.

97 MATHEMATICS AND COMPUTING↗

Short-term solar irradiance forecasting via satellite/model coupling

Here, the short-term (0-3 h) prediction of solar insolation for renewable energy production is a problem well-suited to satellite-based techniques. The spatial, spectral, temporal and radiometric resolution of instrumentation hosted on the geostationary platform allows these satellites to describe the current cloud spatial distribution and optical properties. These properties relate directly to the transient properties of the downwelling solar irradiance at the surface, which come in the form of 'ramps' that pose a central challenge to energy load balancing in a spatially distributed network of solar farms. The short-term evolution of the cloud field may be approximated to first order simply as translational, but care must be taken in how the advection is handled and where the impacts are assigned. In this research, we describe how geostationary satellite observations are used with operational cloud masking and retrieval algorithms, wind field data from Numerical Weather Prediction (NWP), and radiative transfer calculations to produce short-term forecasts of solar insolation for applications in solar power generation. The scheme utilizes retrieved cloud properties to group pixels into contiguous cloud objects whose future positions are predicted using four-dimensional (space + time) model wind fields, selecting steering levels corresponding to the cloud height properties of each cloud group. The shadows associated with these clouds are adjusted for sensor viewing parallax displacement and combined with solar geometry and terrain height to determine the actual location of cloud shadows. For mid/high-level clouds at mid-latitudes and high solar zenith angles, the combined displacements from these geometric considerations are non-negligible. The cloud information is used to initialize a radiative transfer model that computes the direct and diffuse-sky solar insolation at both shadow locations and intervening clear-sky regions. Here, we describe the formulation of the algorithm and validate its performance against Surface Radiation (SURFRAD; Augustine et al., 2000, 2005) network observations. Typical errors range from 8.5% to 17.2% depending on the complexity of cloud regimes, and an operational demonstration outperformed persistence-based forecasting of Global Horizontal Irradiance (GHI) under all conditions by ~10 W/m 2 .

14 SOLAR ENERGY↗

Acceleration of the particle-in-cell code Osiris with graphics processing units

Fully relativistic particle-in-cell (PIC) simulations are crucial for advancing our knowledge of plasma physics. Modern supercomputers based on graphics processing units (GPUs) offer the potential to perform PIC simulations of unprecedented scale, but require robust and feature-rich codes that can fully leverage their computational resources. In this work, this demand is addressed by adding GPU acceleration to the PIC code Osiris. An overview of the algorithm, which features a CUDA extension to the underlying Fortran architecture, is given. Detailed performance benchmarks for thermal plasmas are presented, which demonstrate excellent weak scaling on NERSC's Perlmutter supercomputer and high levels of absolute performance. The robustness of the code to model a variety of physical systems is demonstrated via simulations of Weibel filamentation and laser-wakefield acceleration run with dynamic load balancing. Finally, measurements and analysis of energy consumption are provided that indicate that the GPU algorithm is up to ~14 times faster and ~7 times more energy efficient than the optimized CPU algorithm on a node-to-node basis. The described development addresses the PIC simulation community's computational demands both by contributing a robust and performant GPU-accelerated PIC code and by providing insight into efficient use of GPU hardware.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Implementation of High-Order Multireference Coupled-Cluster Methods on Intel Many Integrated Core Architecture

Here we discuss the implementation of multireference coupled-cluster formalism with singles, doubles, and noniterative triples (MRCCSD(T)), which is capable of taking advantage of the processing power of the Intel Xeon Phi coprocessor. We discuss the integration of two levels of parallelism underlying the MRCCSD(T) implementation with computational kernels designed to offload the computationally intensive parts of the MRCCSD(T) formalism to Intel Xeon Phi coprocessors. Special attention is given to the enhancement of the parallel performance by task reordering that has improved load balancing in the noniterative part of the MRCCSD(T) calculations. We also discuss aspects regarding efficient optimization and vectorization strategies.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Acceleration of Semiempirical QM/MM Methods through Message Passage Interface (MPI), Hybrid MPI/Open Multiprocessing, and Self-Consistent Field Accelerator Implementations

The strategy and implementation of scalable and efficient semiempirical (SE) QM/MM methods in CHARMM are described. The serial version of the code was first profiled to identify routines that required parallelization. Afterward, the code was parallelized and accelerated with three approaches. The first approach was the parallelization of the entire QM/MM routines, including the Fock matrix diagonalization routines, using the CHARMM message passage interface (MPI) machinery. In the second approach, two different self-consistent field (SCF) energy convergence accelerators were implemented using density and Fock matrices as targets for their extrapolations in the SCF procedure. In the third approach, the entire QM/MM and MM energy routines were accelerated by implementing the hybrid MPI/open multiprocessing (OpenMP) model in which both the task- and loop-level parallelization strategies were adopted to balance loads between different OpenMP threads. The present implementation was tested on two solvated enzyme systems (including <100 QM atoms) and an S N 2 symmetric reaction in water. The MPI version exceeded existing SE QM methods in CHARMM, which include the SCC-DFTB and SQUANTUM methods, by at least 4-fold. The use of SCF convergence accelerators further accelerated the code by ~12–35% depending on the size of the QM region and the number of CPU cores used. Although the MPI version displayed good scalability, the performance was diminished for large numbers of MPI processes due to the overhead associated with MPI communications between nodes. This issue was partially overcome by the hybrid MPI/OpenMP approach which displayed a better scalability for a larger number of CPU cores (up to 64 CPUs in the tested systems).

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An Adaptive Fast Multipole Boundary Element Method for Poisson-Boltzmann Electrostatics

The numerical solution of the Poisson Boltzmann (PB) equation is a useful but a computationally demanding tool for studying electrostatic solvation effects in chemical and biomolecular systems. Recently, we have described a boundary integral equation-based PB solver accelerated by a new version of the fast multipole method (FMM). The overall algorithm shows an order N complexity in both the computational cost and memory usage. Here, we present an updated version of the solver by using an adaptive FMM for accelerating the convolution type matrix-vector multiplications. The adaptive algorithm, when compared to our previous nonadaptive one, not only significantly improves the performance of the overall memory usage but also remarkably speeds the calculation because of an improved load balancing between the local- and far-field calculations. We have also implemented a node-patch discretization scheme that leads to a reduction of unknowns by a factor of 2 relative to the constant element method without sacrificing accuracy. As a result of these improvements, the new solver makes the PB calculation truly feasible for large-scale biomolecular systems such as a 30S ribosome molecule even on a typical 2008 desktop computer.

97 MATHEMATICS AND COMPUTING↗

BEYONDPLANCK III. Commander3

We describe the computational infrastructure for end-to-end Bayesian cosmic microwave background (CMB) analysis implemented by the BeyondPlanck Collaboration. The code is called Commander3. It provides a statistically consistent framework for global analysis of CMB and microwave observations and may be useful for a wide range of legacy, current, and future experiments. The paper has three main goals. Firstly, we provide a high-level overview of the existing code base, aiming to guide readers who wish to extend and adapt the code according to their own needs or re-implement it from scratch in a different programming language. Secondly, we discuss some critical computational challenges that arise within any global CMB analysis framework, for instance in-memory compression of time-ordered data, fast Fourier transform optimization, and parallelization and load-balancing. Thirdly, we quantify the CPU and RAM requirements for the current BEYONDPLANCK analysis, finding that a total of 1.5 TB of RAM is required for efficient analysis and that the total cost of a full Gibbs sample for LFI is 170 CPU-hrs, including both low-level processing and high-level component separation, which is well within the capabilities of current low-cost computing facilities. The existing code base is made publicly available under a GNU General Public Library (GPL) license.

79 ASTRONOMY AND ASTROPHYSICS↗

Federated Access from DOE Labs to Distributed Storage in the EIC Era of Computing

The Electron Ion Collider (EIC) collaboration and future experiment is a unique scientific ecosystem within Nuclear Physics as the experiment starts right off as a crosscollaboration between Brookhaven National Lab (BNL) & Jefferson Lab (JLab). As a result, this muti-lab computing model tries at best to provide services accessible from anywhere by anyone who is part of the collaboration. While the computing model for the EIC is not finalized, it is anticipated that the computational and storage resources will be made accessible to a wide range of collaborators across the world. The use of federated ID seems to be a critical element to the strategy of providing such services, allowing seamless access to each lab site computing resources. However, providing Federated access to a Federated storage is not a trivial matter and has its share of technical challenges. In this contribution, we focus on the steps we took towards the deployment of a distributed object storage system that integrates with Amazon S3 and Federated ID. We will first cover for and explain the first stage storage solutions provided to the EIC during the detector design phase. Our initial test deployment consisted of Lustre storage using MinIO, hence providing an S3 interface. High Availability load balancers were added later to provide the initial scalability it lacked. Performance of that system will be shown. While this embryonic solution worked well, it had many limitations. Looking ahead, the Ceph object storage is considered a top-of-the-line solution in the storage community - since the Ceph Object Gateway is compatible with the Amazon S3 API out of the box, our next phase will use a native S3 storage. Our Ceph deployment will consist of erasure coded storage nodes to maximize storage potential along with multiple Ceph Object Gateways for redundant access. We will compare performance of our next stage implementations. Finally, we will present how to leverage OpenID Connect with the Ceph Object Gateway’s to enable Federated ID access. We hope this contribution will serve the community needs as we move forward with cross-lab collaborations and the need for Federated ID access to distributed compute facilities.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Modeling Cardiac Electrophysiology at the Organ Level in the Peta FLOPS Computing Age

Despite a steep increase in available compute power, in-silico experimentation with highly detailed models of the heart remains to be challenging due to the high computational cost involved. It is hoped that next generation high performance computing (HPC) resources lead to significant reductions in execution times to leverage a new class of in-silico applications. However, performance gains with these new platforms can only be achieved by engaging a much larger number of compute cores, necessitating strongly scalable numerical techniques. So far strong scalability has been demonstrated only for a moderate number of cores, orders of magnitude below the range required to achieve the desired performance boost.In this study, strong scalability of currently used techniques to solve the bidomain equations is investigated. Benchmark results suggest that scalability is limited to 512-4096 cores within the range of relevant problem sizes even when systems are carefully load-balanced and advanced IO strategies are employed.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The development of laser-plasma interaction program LAP3D on thousands of processors

Modeling laser-plasma interaction (LPI) processes in real-size experiments scale is recognized as a challenging task. For explorering the influence of various instabilities in LPI processes, a three-dimensional laser and plasma code (LAP3D) has been developed, which includes filamentation, stimulated Brillouin backscattering (SBS), stimulated Raman backscattering (SRS), non-local heat transport and plasmas flow computation modules. In this program, a second-order upwind scheme is applied to solve the plasma equations which are represented by an Euler fluid model. Operator splitting method is used for solving the equations of the light wave propagation, where the Fast Fourier translation (FFT) is applied to compute the diffraction operator and the coordinate translations is used to solve the acoustic wave equation. The coupled terms of the different physics processes are computed by the second-order interpolations algorithm. In order to simulate the LPI processes in massively parallel computers well, several parallel techniques are used, such as the coupled parallel algorithm of FFT and fluid numerical computation, the load balance algorithm, and the data transfer algorithm. Now the phenomena of filamentation, SBS and SRS have been studied in low-density plasma successfully with LAP3D. Scalability of the program is demonstrated with a parallel efficiency above 50% on about ten thousand of processors.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Multi-level parallelization of quantum-chemical calculations

Here, strategies for multiple-level parallelizations of quantum-mechanical calculations are discussed, with an emphasis on using groups of workers for performing parallel tasks. These parallel programming models can be used for a variety ab initio quantum chemistry approaches, including the fragment molecular orbital method and replica-exchange molecular dynamics. Strategies for efficient load balancing on problems of increasing granularity are introduced and discussed. A four-level parallelization is developed based on a multi-level hierarchical grouping, and a high parallel efficiency is achieved on the Theta supercomputer using 131 072 OpenMP threads.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A parallel, distributed memory implementation of the adaptive sampling configuration interaction method

The many-body simulation of quantum systems is an active field of research that involves several different methods targeting various computing platforms. Many methods commonly employed, particularly coupled cluster methods, have been adapted to leverage the latest advances in modern high-performance computing. Selected configuration interaction (sCI) methods have seen extensive usage and development in recent years. However, the development of sCI methods targeting massively parallel resources has been explored only in a few research works. Here, we present a parallel, distributed memory implementation of the adaptive sampling configuration interaction approach (ASCI) for sCI. In particular, we will address the key concerns pertaining to the parallelization of the determinant search and selection, Hamiltonian formation, and the variational eigenvalue calculation for the ASCI method. Load balancing in the search step is achieved through the application of memory-efficient determinant constraints originally developed for the ASCI-PT2 method. The presented benchmarks demonstrate near optimal speedup for ASCI calculations of Cr 2 (24e, 30o) with 10 6 , 10 7 , and 3 × 10 8 variational determinants on up to 16 384 CPUs. Importantly, to the best of the authors’ knowledge, this is the largest variational ASCI calculation to date.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Cryogenic aspects of a 20 MW class low-temperature superconducting generator for the renewables industry

In this paper we give a progress update on the cryogenic design for cooling a 20 MW class, partially superconducting generator with stationary field coils for offshore wind renewables industry [1-2]. This is a continuation of an earlier program on a 10 MW system dating back ten years. Whereas this new power rating increase leads to a radial diameter expansion of the superconducting field coils from 4 to > 9.5 m, it maintains the axial length. We show the design based on this scaled up with enlarged diameter. The field coil size increase asks for higher cooling power that leads to a bigger cold box size to accommodate the cryocoolers. In the design, as many as 8 cryocoolers can be accessed and serviced from the nacelle that houses the main cryogenic components. A typical thermal load balance sheet with all components is given and compared against the available cryocooler cooling power at different operating conditions. Due to the cryocoolers’ local point of contact cooling within the nacelle, the temperature gradient of the thermal shield that fully encloses the cold mass with its embedded field coils needs to be balanced out to minimize the heat burden on the field coils. This requires additional analytical efforts and implementation of further design features. The extended nacelle houses the cold box with its cryogenic infrastructure. The interface for the envisaged cryogenic pushbutton closed-loop circulating system remains invisible, requires no handling of cryogenic liquids and is hermetically closed. The field coil diameter increase also leads to a greater initial helium gas storage volume. In addition, it requires higher initial room temperature fill pressure within the toroidal helium vapor storage tanks and in cooling tubes connecting to the cryocoolers. The toroidal helium gas tanks are thermally coupled with the thermal shield so that helium convection inside the storage tank can improve heat transfer and reduce the temperature gradient of the thermal shield. Besides those heat transfer challenges, additional mechanical strain within this large structure is exerted on the torque tubes during initial cooldown and when energizing field coils. Some of those design challenges are quite unexpected, leading to novel workarounds in order to maintain the chosen cooling strategy. Finally, we assess those design limitations in view of further cryogenic scalability with emphasis on manufacturability and assembly.

cryogenics↗

COMPASO : A new halo finder for competitive assignment to spherical overdensities

We describe a new method (compaso) for identifying groups of particles in cosmological N-body simulations. compaso builds upon existing spherical overdensity (SO) algorithms by taking into consideration the tidal radius around a smaller halo before competitively assigning halo membership to the particles. In this way, the compaso finder allows for more effective deblending of haloes in close proximity as well as the formation of new haloes on the outskirts of larger ones. This halo-finding algorithm is used in the abacussummit suite of N-body simulations, designed to meet the cosmological simulation requirements of the Dark Energy Spectroscopic Instrument (DESI) survey. compaso is developed as a highly efficient on-the-fly group finder, which is crucial for enabling good load-balancing between the GPU and CPU and the creation of high-resolution merger trees. In this paper, we describe the halo-finding procedure and its particular implementation in abacus, accompanying it with a qualitative analysis of the finder. We test the robustness of the compaso catalogues before and after applying the cleaning method described in an accompanying paper and demonstrate its effectiveness by comparing it with other validation techniques. We then visualize the haloes and their density profiles, finding that they are well fit by the NFW formalism. Finally, we compare other properties such as radius–mass relationships and two-point correlation functions with that of another widely used halo finder, rockstar.

79 ASTRONOMY AND ASTROPHYSICS↗

Scalable and fast heterogeneous molecular simulation with predictive parallelization schemes

Multiscale and inhomogeneous molecular systems are challenging topics in the field of molecular simulation. In particular, modeling biological systems in the context of multiscale simulations and exploring material properties are driving a permanent development of new simulation methods and optimization algorithms. In computational terms, those methods require parallelization schemes that make a productive use of computational resources for each simulation and from its genesis. Here, we introduce the heterogeneous domain decomposition approach, which is a combination of an heterogeneity-sensitive spatial domain decomposition with an a priori rearrangement of subdomain walls. Within this approach and paper, the theoretical modeling and scaling laws for the force computation time are proposed and studied as a function of the number of particles and the spatial resolution ratio. We also show the new approach capabilities, by comparing it to both static domain decomposition algorithms and dynamic load-balancing schemes. Specifically, two representative molecular systems have been simulated and compared to the heterogeneous domain decomposition proposed in this work. Finally, these two systems comprise an adaptive resolution simulation of a biomolecule solvated in water and a phase-separated binary Lennard-Jones fluid.

36 MATERIALS SCIENCE↗

LP-BNN: Ultra-low-Latency BNN Inference with Layer Parallelism

High inference latency seriously limits the deployment of DNNs in real-time domains such as autonomous driving, robotic control, and many others. To address this emerging challenge, researchers have proposed approximate DNNs with reduced precision, e.g., Binarized Neural Networks (BNNs). While BNNs can be built to have little loss in accuracy, latency reduction still has much room for improvement. In this paper, we propose a single-FPGA-based BNN accelerator that achieves microsecond-level ultra-low-latency inference of ImageNet, LP- BNN. We obtain this performance via several design optimizations. First, we optimize the network structure by removing Batch Normalization (BN) functions which leads to significant latency in BNNs without any loss on accuracy. Second, we propose a parameterized architecture which is based on layer parallelism and supports nearly perfect load balancing for various types of BNNs. Third, we fuse all the convolution layers and the first fully connected layer. We process them in parallel through fine- grained inter-layer pipelining. With our proposed accelerator, the inference of binarized AlexNet, VGGNet, and ResNet are completed within 21.5µs, 355µs, and 67.8µs respectively, with no loss in accuracy as compared with other BNN implementations.

Geng, Tong↗

Accelerating semantic graph databases on commodity clusters

We are developing a full software system for accelerating semantic graph databases on commodity cluster that scales to hundreds of nodes while maintaining constant query throughput. Our framework comprises a SPARQL to C++ compiler, a library of parallel graph methods and a custom multithreaded runtime layer, which provides a Partitioned Global Address Space (PGAS) programming model with fork/join parallelism and automatic load balancing over a commodity clusters. We present preliminary results for the compiler and for the runtime.

commodity clusters↗

Large Scale Parallel Solution of Incompressible Flow Problems Using Uintah and Hypre

The Uintah Software framework was developed to provide an environment for solving fluid-structure interaction problems on structured adaptive grids on large-scale, long-running, data-intensive problems. Uintah uses a combination of fluid-flow solvers and particle-based methods for solids together with a novel asynchronous task-based approach with fully automated load balancing. As Uintah is often used to solve incompressible flow problems in combustion applications it is important to have a scalable linear solver. While there are many such solvers available, the scalability of those codes varies greatly. The hypre software offers a range of solvers and pre-conditioners for different types of grids. The weak scalability of Uintah and hypre is addressed for particular examples of both packages when applied to a number of incompressible flow problems. After careful software engineering to reduce startup costs, much better than expected weak scalability is seen for up to 100K cores on NSFs Kraken architecture and up to260K cpu cores, on DOEs new Titan machine. The scalability is found to depend in a crtitical way on the choice of algorithm used by hypre for a realistic application problem.

Computer Science↗