Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “High Performance Computing (HPC)”

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 1,387 records · Page 77

OpenSHMEM-UCX: Evaluation of UCX for Implementing OpenSHMEM Programming Model, In: OpenSHMEM 2016: OpenSHMEM and Related Technologies. Enhancing OpenSHMEM for Hybrid Environments

The OpenSHMEM reference implementation was developed towards the goal of developing an open source and high-performing OpenSHMEM implementation. To achieve portability and performance across various networks, the OpenSHMEM reference implementation uses GASNet and UCCS for network operations. Recently, new network layers have emerged with the promise of providing high-performance, scalability, and portability for HPC applications. In this paper, we implement the OpenSHMEM reference implementation to use the UCX framework for network operations. Then, we evaluate its performance and scalability on Cray XK systems to understand UCX’s suitability for developing the OpenSHMEM programming model. Further, we develop a benchmark called SHOMS for evaluating the OpenSHMEM implementation. Our experimental results show that OpenSHMEM-UCX outperforms the vendor supplied OpenSHMEM implementation in most cases on the Cray XK system by up to 40% with respect to message rate and up to 70% for the execution of application kernels.

Computer Science↗

HPC Programming on Intel Many-Integrated-Core Hardware with MAGMA Port to Xeon Phi

This paper presents the design and implementation of several fundamental dense linear algebra (DLA) algorithms for multicore with Intel Xeon Phi coprocessors. In particular, we consider algorithms for solving linear systems. Further, we give an overview of the MAGMA MIC library, an open source, high performance library, that incorporates the developments presented here and, more broadly, provides the DLA functionality equivalent to that of the popular LAPACK library while targeting heterogeneous architectures that feature a mix of multicore CPUs and coprocessors. The LAPACK-compliance simplifies the use of the MAGMA MIC library in applications, while providing them with portably performant DLA. High performance is obtained through the use of the high-performance BLAS, hardware-specific tuning, and a hybridization methodology whereby we split the algorithm into computational tasks of various granularities. Execution of those tasks is properly scheduled over the heterogeneous hardware by minimizing data movements and mapping algorithmic requirements to the architectural strengths of the various heterogeneous hardware components. Our methodology and programming techniques are incorporated into the MAGMA MIC API, which abstracts the application developer from the specifics of the Xeon Phi architecture and is therefore applicable to algorithms beyond the scope of DLA.

97 MATHEMATICS AND COMPUTING↗

Integrating Distributed-Memory Machine Learning into Large-Scale HPC Simulations

Lawrence Livermore National Laboratory (LLNL) is a national lab based in Livermore, California. It was established in 1952 to meet increasing national security needs. They specialize in earth and atmospheric science, bioscience and bioengineering, lasers and optical science and technology, advanced materials and manufacturing, high-energy-density science, nuclear, chemical, and isotopic science and technology, and HPC, simulation and data science ((cor, 2018)). LLNL is also home to a number of large Linux clusters and supercomputers which are used to support research in their areas of specialization (mac, 2018). One purpose of the LLNL supercomputers is to run highly detailed and accurate physics simulations. These simulations allow scientists to get data on experiments that would be too expensive, difficult, or dangerous to perform in reality. With the help of simulations scientists at LLNL are able to study processes in more depth than they would be able to otherwise.

97 MATHEMATICS AND COMPUTING↗

Toward an optimal online checkpoint solution under a two-level HPC checkpoint model

The traditional single-level checkpointing method suffers from significant overhead on large-scale platforms. Hence, multilevel checkpointing protocols have been studied extensively in recent years. The multilevel checkpoint approach allows different levels of checkpoints to be set (each with different checkpoint overheads and recovery abilities), in order to further improve the fault tolerance performance of extreme-scale HPC applications. How to optimize the checkpoint intervals for each level, however, is an extremely difficult problem. In this paper, we construct an easy-to-use two-level checkpoint model. Checkpoint level 1 deals with errors with low checkpoint/recovery overheads such as transient memory errors, while checkpoint level 2 deals with hardware crashes such as node failures. Compared with previous optimization work, our new optimal checkpoint solution offers two improvements: (1) it is an online solution without requiring knowledge of the job length in advance, and (2) it shows that periodic patterns are optimal and determines the best pattern. We evaluate the proposed solution and compare it with the most up-to-date related approaches on an extreme-scale simulation testbed constructed based on a real HPC application execution. Simulation results show that our proposed solution outperforms other optimized solutions and can improve the performance significantly in some cases. Specifically, with the new solution the wall-clock time can be reduced by up to 25.3% over that of other state-of-the-art approaches. Lastly, a brute-force comparison with all possible patterns shows that our solution is always within 1% of the best pattern in the experiments.

97 MATHEMATICS AND COMPUTING↗

HPC application fault-tolerance using transparent redundant computation.

As the core count of HPC machines continue to grow in size, issues such as fault tolerance and reliability are becoming limiting factors for application scalability. Current techniques to ensure progress across faults, for example coordinated checkpoint-restart, are unsuitable for machines of this scale due to their predicted high overheads. In this study, we present the design and implementation of a novel system for ensuring reliability which uses transparent, rank-level, redundant computation. Using this system, we show the overheads involved in redundant computation for a number of real-world HPC applications. Additionally, we relate the communication characteristics of an application to the overheads observed.

97 MATHEMATICS AND COMPUTING↗

Dynamic Undervolting to Improve Energy Efficiency on Multicore X86 CPUs

Chip manufacturers introduce redundancy at various levels of CPU design to guarantee correct operation, even for worst-case combinations of non-idealities in process variation and system operating conditions. This redundancy is implemented partly in the form of voltage margins. However, for a wide range of real-world execution scenarios these margins are excessive and merely translate to increased power consumption, hindering the effort towards higher-energy efficiency in both HPC and general purpose computing. Our study on the x86-64 Haswell and Skylake multicore microarchitectures reveals-wide voltage margins, which vary across different microarchitectures, different chip parts of the same microarchitecture, and across different workloads. We find that it is necessary to quantify-voltage margins using multi-threaded and multi-instance workloads, as characterization with single-threaded and single-instance workloads that do not stress the CPU to its full capacity typically identifies overly optimistic margins that lead to errors when applied in realistic program execution scenarios. In addition, we introduce, deploy and evaluate a run-time governor that dynamically reduces the supply voltage of modern multicore x86-64 CPUs. Our governor employs a model that takes as input a set of performance metrics which are directly measurable via performance monitoring counters and have high predictive value for the minimum tolerable supply voltage (V min ), to predict and apply the appropriate reduction for the workload at hand. Compared with the conventional DVFS governor, our approach in this study achieves up to 42 percent energy savings for the Skylake family and 34 percent for the Haswell family for complex, real-world applications.

97 MATHEMATICS AND COMPUTING↗

Continuous whole-system monitoring toward rapid understanding of production HPC applications and systems

A detailed understanding of HPC applications’ resource needs and their complex interactions with each other and HPC platform resources are critical to achieving scalability and performance. Such understanding has been difficult to achieve because typical application profiling tools do not capture the behaviors of codes under the potentially wide spectrum of actual production conditions and because typical monitoring tools do not capture system resource usage information with high enough fidelity to gain sufficient insight into application performance and demands. In this paper we present both system and application profiling results based on data obtained through synchronized system wide monitoring on a production HPC cluster at Sandia National Laboratories (SNL). We demonstrate analytic and visualization techniques that we are using to characterize application and system resource usage under production conditions for better understanding of application resource needs. Furthermore, our goals are to improve application performance (through understanding application-to-resource mapping and system throughput) and to ensure that future system capabilities match their intended workloads.

97 MATHEMATICS AND COMPUTING↗

Data-intensive management and analysis for scientifc simulations.

Scientific simulations can produce enormous amounts of data, making the analysis of results and management of files a difficult task for scientists. The simulation management and analysis system (Smaash) described here is designed to allow scientists to easily capture, store, organize, monitor, and analyze simulation results. The system is automatic, standardized, and secure. Smaash was built using open-source tools and modularized to be independent of the scientific simulation. The web-based front-end allows the scientist to easily interact with the data, and has proved its usefulness in improving the efficiency of a scientific team's workflow. High performance parallel computing allows scientists to solve complex physical problems through computer simulation. However, the massive amounts of data generated and the complex computing environment can create additional complications. A recent review by Ludaescher et al.(2009) describes how scientific workflows can assist scientists in extracting knowledge from these data-intensive operations by automating components within pipelines. Within the fusion community, Klasky et al.(2008) and colleagues have developed a system that handles the storage management, data movement, metadata generation and management, and a means to analyze the results. In response to scientists needs, a simulation management and analysis system (Smaash) was developed at the University of Chicago and Argonne National Laboratory (USA). Smaash provides an integrated way to monitor simulations and analyze computational results; catalog, store, and retrieve simulations; and prepare output for publications. The system is independent of the particular simulation code, accessible from many HPC and browser-based platforms, and built around open-source software tools. Data security and provenance is considered throughout. The analysis components are hidden behind a web-based front end, enabling scientists to focus on their results and not get bogged down by information overload.

99 GENERAL AND MISCELLANEOUS↗

Program Development Tools and Infrastructures: Activities by ASC CSSE/CCE and Partners

Exascale class machines will exhibit a new level of complexity: they will feature an unprecedented number of cores and threads, will most likely be heterogeneous and deeply hierarchical, and offer a range of new hardware techniques (such as speculative threading, transactional memory, programmable prefetching, and programmable accelerators), which all have to be utilized for an application to realize the full potential of the machine. Additionally, users will be faced with less memory per core, fixed total power budgets, and sharply reduced MTBFs. At the same time, it is expected that the complexity of applications will rise sharply for exascale systems, both to implement new science possible at exascale and to exploit the new hardware features necessary to achieve exascale performance. This is particularly true for many of the NNSA codes, which are large and often highly complex integrated simulation codes that push the limits of everything in the system including language features. To overcome these limitations and to enable users to reach exascale performance, users will expect a new generation of tools that address the bottlenecks of exascale machines, that work seamlessly with the (set of) programming models on the target machines, that scale with the machine, that provide automatic analysis capabilities, and that are flexible and modular enough to overcome the complexities and changing demands of the exascale architectures. Further, any tool must be robust enough to handle the complexity of large integrated codes while keeping the user's learning curve low. With the ASC program, in particular the CSSE (Computational Systems and Software Engineering) and CCE (Common Compute Environment) projects, we are working towards a new generation of tools that fulfill these requirements and that provide our users as well as the larger HPC community with the necessary tools, techniques, and methodologies required to make exascale performance a reality.

43 PARTICLE ACCELERATORS↗

Comparison of Resource Platform Selection Approaches for Scientific Workflows

Cloud computing is increasingly considered as an additional computational resource platform for scientific workflows. The cloud offers opportunity to scale-out applications from desktops and local cluster resources. At the same time, it can eliminate the challenges of restricted software environments and queue delays in shared high performance computing environments. Choosing from these diverse resource platforms for a workflow execution poses a challenge for many scientists. Scientists are often faced with deciding resource platform selection trade-offs with limited information on the actual workflows. While many workflow planning methods have explored task scheduling onto different resources, these methods often require fine-scale characterization of the workflow that is onerous for a scientist. In this position paper, we describe our early exploratory work into using blackbox characteristics to do a cost-benefit analysis across of using cloud platforms. We use only very limited high-level information on the workflow length, width, and data sizes. The length and width are indicative of the workflow duration and parallelism. The data size characterizes the IO requirements. We compare the effectiveness of this approach to other resource selection models using two exemplar scientific workflows scheduled on desktops, local clusters, HPC centers, and clouds. Early results suggest that the blackbox model often makes the same resource selections as a more fine-grained whitebox model. We believe the simplicity of the blackbox model can help inform a scientist on the applicability of cloud computing resources even before porting an existing workflow.

97 MATHEMATICS AND COMPUTING↗

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↗

Evaluation of pre-training large language models on leadership-class supercomputers

Large language models (LLMs) have arisen rapidly to the center stage of artificial intelligence as the foundation models applicable to many downstream learning tasks. However, how to effectively build, train, and serve such models for many high-stake and first-principle-based scientific use cases are both of great interests and of great challenges. Moreover, pre-training LLMs with billions or even trillions of parameters can be prohibitively expensive not just for academic institutions, but also for well-funded industrial and government labs. Furthermore, the energy cost and the environmental impact of developing LLMs must be kept in mind. Here, in this work, we conduct a first-of-its-kind performance analysis to understand the time and energy cost of pre-training LLMs on the Department of Energy (DOE)’s leadership-class supercomputers. Employing state-of-the-art distributed training techniques, we evaluate the computational performance of various parallelization approaches at scale for a range of model sizes, and establish a projection model for the cost of full training. Our findings provide baseline results, best practices, and heuristics for pre-training such large models that should be valuable to HPC community at large. We also offer insights and optimization strategies for using the first exascale computing system, Frontier, to train models of the size of GPT-3 and beyond.

97 MATHEMATICS AND COMPUTING↗

Optimizing Asynchronous Multi-Level Checkpoint/Restart Configurations with Machine Learning

With the emergence of versatile storage systems, multi-level checkpointing (MLC) has become a common approach to gain efficiency. However, multi-level checkpoint/restart can cause enormous I/O traffic on HPC systems. To use multilevel checkpointing efficiently, it is important to optimize check-point/restart configurations. Current approaches, namely modeling and simulation, are either inaccurate or slow in determining the optimal configuration for a large scale system. In this paper, we show that machine learning models can be used in combination with accurate simulation to determine the optimal checkpoint configurations. We also demonstrate that more advanced techniques such as neural networks can further improve the performance in optimizing checkpoint configurations.

checkpoint-restart↗

Integrating and Characterizing HPC Task Runtime Systems for hybrid AI-HPC workloads

Scientific workflows increasingly involve both HPC and machine-learning tasks, combining MPI-based simulations, training, and inference in a single execution. Launchers such as Slurm’s srun constrain concurrency and throughput, making them unsuitable for dynamic and heterogeneous workloads. We present a performance study of RADICAL-Pilot (RP) integrated with Flux and Dragon, two complementary runtime systems that enable hierarchical resource management and high-throughput function execution. Using synthetic and production-scale workloads on Frontier, we characterize the task execution properties of RP across runtime configurations. RP+Flux sustains up to 930 tasks/s, and RP+Flux+Dragon exceeds 1,500 tasks/s with over 99.6% utilization. In contrast, srun peaks at 152 tasks/s and degrades with scale, with utilization below 50%. For IMPECCABLE.v2 drug discovery campaign, RP+Flux reduces makespan by 30–60% relative to srun/Slurm and increases throughput more than four times on up to 1,024. These results demonstrate hybrid runtime integration in RP as a scalable approach for hybrid AI-HPC workloads.

HPC-AI↗

Evaluation of Simple Causal Message Logging for Large-Scale Fault Tolerant HPC Systems

The era of petascale computing brought machines with hundreds of thousands of processors. The next generation of exascale supercomputers will make available clusters with millions of processors. In those machines, mean time between failures will range from a few minutes to few tens of minutes, making the crash of a processor the common case, instead of a rarity. Parallel applications running on those large machines will need to simultaneously survive crashes and maintain high productivity. To achieve that, fault tolerance techniques will have to go beyond checkpoint/restart, which requires all processors to roll back in case of a failure. Incorporating some form of message logging will provide a framework where only a subset of processors are rolled back after a crash. In this paper, we discuss why a simple causal message logging protocol seems a promising alternative to provide fault tolerance in large supercomputers. As opposed to pessimistic message logging, it has low latency overhead, especially in collective communication operations. Besides, it saves messages when more than one thread is running per processor. Finally, we demonstrate that a simple causal message logging protocol has a faster recovery and a low performance penalty when compared to checkpoint/restart. Running NAS Parallel Benchmarks (CG, MG and BT) on 1024 processors, simple causal message logging has a latency overhead below 5%.

97 MATHEMATICS AND COMPUTING↗

Photoelectron spectroscopy and computational investigations of the electronic structures and noncovalent interactions of cyclodextrin-closo-dodecaborate anion complexes x-CD·B12X122- (x = a, ß, y; X = H, F)

We report a joint negative ion photoelectron spectroscopy (NIPES) and computational study on the electronic structures and noncovalent interactions of a series of cyclodextrin-closo-dodecaborate dianion complexes, ?-CD·B12X122- (? = a, ß, ?; X = H, F). The measured vertical / adiabatic detachment energies (VDEs / ADEs) are 1.15/0.93, 3.55/3.20, 3.90/3.60, and 3.85/3.60 eV for B12H122- and its a-, ß-, ?-CD complexes, respectively; while the corresponding values are 1.90/1.70, 4.00/3.60, 4.33/3.95, and 4.30/3.85 eV for the X = F case. These results show that the inclusion of B12X122- into the CD cavities greatly increase the electronic stability of the dianions. The effect of electronic stabilization for ß-CD is roughly the same as for ?-CD, both being considerably stronger than that for a-CD. Density functional theory (DFT) based geometry optimization reveals that B12X122- are inserted into CDs increasingly deeper from a-CD to ?-CD. The calculated VDEs and ADEs agree with the experiments well, particularly, reproducing the electron binding energy (EBE) trends. The molecular orbital analyses indicate that the most loosely bound photodetached electrons origin from the guest B12X122- moieties. In addition to a shift of all signals to larger EBE, significant changes in the signal patterns are observed. At low EBE, this is due to the splitting of highly degenerate B12X122- orbitals, while at high EBE, photodetachment from CD oxygens contributes to the new bands. The guest B12X122- and host CD nocovalent, size-specific interaction based on the independent gradient model (IGM) and energy decomposition analysis (EDA), is dominated by electrostatic interactions. The analysis further unravels unambiguiously the existence of dihydrogen bonding and how it affects the total energy that stabilizes the host-guest complexes of CDs·B12H122- compared to the general hydrogen bonding interaction in CDs·B12F122-. This work clearly exhibits strong influences on the electronic structures of dodecaborates upon clustering with CDs, with both size (a-, ß-, ?-) and molecular (X = H or F) specificities, thus providing critical molecular-level information on the cyclodextrin-closo-dodecaborate interactions of interest to medical applications, e.g. Boron neutron capature therapy. The NIPES experiments done at PNNL were supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences and Bioscience (X.-B.W.) and was performed at the EMSL, a national scientific user facility sponsored by DOE’s Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory. H.S. and Z.S. acknowledge the funding support of National Natural Science Foundation of China (Nos.11727810, 61720106009 and 21603074), the Science and Technology Commission of Shanghai Municipality (Nos. 19JC1412200), and the Program of Introducing Talents of Discipline to Universities 111 project (B12024). Z. L. thanks the China Scholarship Council (CSC) for financial support. We acknowledge the ECNU Multifunctional Platform for Innovation (001) and HPC Research Computing Team for providing computational and storage resources. J.W. acknowledges support from the Alexander von Humboldt foundation (Feodor Lynen Fellowship and Rückkehrerstipendium), a Freigeist fellowship of the Volkswagenfoundation and Prof. Vladimir A. Azov for helpful discussions.

Li, Zhipeng↗

Indicator-directed Dynamic Power Management for Iterative Workloads on GPU-Accelerated Systems

Modern high-performance and warehouse computing centers show strong interest in minimizing system power consumption while satisfying customers’ quality of service (QoS). Dynamic voltage and frequency scaling (DVFS) is effective for achieving this goal. Nevertheless, automating the process online and making it transparent to users must address three major challenges: (1) Complexity — today’s hardware components (e.g., CPUs, GPUs, memory, network, etc.) can be configured in several or dozens of frequency/voltage states for satisfying divergent system demands. Given their combination and the emergence of heterogeneity, searching the optimal configuration in the design space online can be timing consuming. (2) QoS guarantee — user-defined objectives such as power constraint and performance target must be monitored, predicted and ensured at the best effort. (3) Adaptability — various known and unknown workloads run on systems. Workloads characteristics should be quickly determined and configurations dynamically adjusted in accord with workloads and QoS. In this work, we focus on applications exhibiting an interesting feature – iterative or periodic, which is common among conventional HPC and emerging machine learning workloads. We propose an online dynamic power-performance (ODPP) management framework to dynamically adjust GPU DVFS configurations to meet performance and power objectives and constraints, without any code annotation or intrusion. Particularly, ODPP extracts the performance and power indicators for applications from their resources utilization profiles in a short episode. It further automatically constructs an accurate model that infers from the indicators how the application's performance and power vary with GPU core and memory frequencies. Aided with the model, for both seen and unseen applications, ODPP can quickly determine the most appropriate DVFS configuration for their execution. We evaluate ODPP on an NVIDIA GPU using multiple exascale computing (ECP) and deep learning applications.

Zou, Pengfei↗

A Highly Parallel Implementation of K-Means for Multithreaded Architecture

We present a parallel implementation of the popular k-means clustering algorithm for massively multithreaded computer systems, as well as a parallelized version of the KKZ seed selection algorithm. We demonstrate that as system size increases, sequential seed selection can become a bottleneck. We also present an early attempt at parallelizing k-means that highlights critical performance issues when programming massively multithreaded systems. For our case studies, we used data collected from electric power simulations and run on the Cray XMT.

99 GENERAL AND MISCELLANEOUS↗