Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “asynchronous processing”

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 253 records · Page 14

On the Suitability of MPI as a PGAS Runtime

Partitioned Global Address Space (PGAS) models are emerging as a popular alternative to MPI models for designing scalable applications. At the same time, MPI remains a ubiquitous communication subsystem due to its standardization, high performance, and availability on leading platforms. In this paper, we explore the suitability of using MPI as a scalable PGAS communication subsystem. We focus on the Remote Memory Access (RMA) communication in PGAS models which typically includes {\em get, put,} and {\em atomic memory operations}. We perform an in-depth exploration of design alternatives based on MPI. These alternatives include using a semantically-matching interface such as MPI-RMA, as well as not-so-intuitive interfaces such as MPI two-sided with a combination of multi-threading and dynamic process management. With an in-depth exploration of these alternatives and their shortcomings, we propose a novel design which is facilitated by the data-centric view in PGAS models. This design leverages a combination of highly tuned MPI two-sided semantics and an automatic, user-transparent split of MPI communicators to provide asynchronous progress. We implement the asynchronous progress ranks approach and other approaches within the Communication Runtime for Exascale which is a communication subsystem for Global Arrays. Our performance evaluation spans pure communication benchmarks, graph community detection and sparse matrix-vector multiplication kernels, and a computational chemistry application. The utility of our proposed PR-based approach is demonstrated by a 2.17x speed-up on 1008 processors over the other MPI-based designs.

Computational modeling↗

The Meteorological Measurement System on the NASA ER-2 aircraft

A Meteorological Measurement System (MMS) was designed and installed on one of the NASA high-altitude ER-2 aircraft (NASA 706). The MMS provides in situ measurements of free-stream pressure (+ or - 0.3 mb), temperature (+ or - 0.3 C), and wind vector (+ or - 1 m/s). It incorporates a high-resolution inertial navigation system specially configured for scientific applications, a radome differential pressure system for measurements of the airflow angles, and a compact, computer-controlled data acquisition system to sample, process and store 45 variables on tape and on disk. The MMS hardware and software development is described, and resolution and accuracy of the instrumentation discussed. Custom software facilitates preflight system checkout, inflight data acquisition, and fast postflight data download. It accommodates various modes of MMS data: analog and digital, serial and parallel, and synchronous and asynchronous. Flight results are presented to demonstrate the capability of the system.

Scott, Stan G.↗

Unraveling Network-induced Memory Contention: Deeper Insights with Machine Learning

Remote Direct Memory Access (RDMA) is expected to be an integral communication mechanism for future exascale systems enabling asynchronous data transfers, so that applications may fully utilize CPU resources while simultaneously sharing data amongst remote nodes. We examine Network-induced Memory Contention (NiMC) on Infiniband networks. We expose the interactions between RDMA, main-memory and cache, when applications and out-of-band services compete for memory resources. We then explore NiMCs resulting impact on application-level performance. For a range of hardware technologies and HPC workloads, we quantify NiMC and show that NiMCs impact grows with scale resulting in up to 3X performance degradation at scales as small as 8K processes even in applications that previously have been shown to be performance resilient in the presence of noise. In addition, this work examines the problem of predicting NiMC's impact on applications by leveraging machine learning and easily accessible performance counters. This approach provides additional insights about the root cause of NiMC and facilitates dynamic selection of potential solutions. Finally, we evaluated three potential techniques to reduce NiMCs impact, namely hardware offloading, core reservation and network throttling.

97 MATHEMATICS AND COMPUTING↗

Scaling Deep Learning on GPU and Knights Landing clusters

The speed of deep neural networks training has become a big bottleneck of deep learning research and development. For example, training GoogleNet by ImageNet dataset on one Nvidia K20 GPU needs 21 days. To speed up the training process, the current deep learning systems heavily rely on the hardware accelerators. However, these accelerators have limited on-chip memory compared with CPUs. To handle large datasets, they need to fetch data from either CPU memory or remote processors. We use both self-hosted Intel Knights Landing (KNL) clusters and multi-GPU clusters as our target platforms. From an algorithm aspect, current distributed machine learning systems are mainly designed for cloud systems. These methods are asynchronous because of the slow network and high fault-tolerance requirement on cloud systems. We focus on Elastic Averaging SGD (EASGD) to design algorithms for HPC clusters. Original EASGD used round-robin method for communication and updating. The communication is ordered by the machine rank ID, which is inefficient on HPC clusters. First, we redesign four efficient algorithms for HPC systems to improve EASGD's poor scaling on clusters. Async EASGD, Async MEASGD, and Hogwild EASGD are faster \textcolor{black}{than} their existing counterparts (Async SGD, Async MSGD, and Hogwild SGD, resp.) in all the comparisons. Finally, we design Sync EASGD, which ties for the best performance among all the methods while being deterministic. In addition to the algorithmic improvements, we use some system-algorithm codesign techniques to scale up the algorithms. By reducing the percentage of communication from 87% to 14%, our Sync EASGD achieves 5.3x speedup over original EASGD on the same platform. We get 91.5% weak scaling efficiency on 4253 KNL cores, which is higher than the state-of-the-art implementation.

60 APPLIED LIFE SCIENCES↗

HPCTOOLKIT: tools for performance analysis of optimized parallel programs

HPCTOOLKIT is an integrated suite of tools that supports measurement, analysis, attribution, and presentation of application performance for both sequential and parallel programs. HPCTOOLKIT can pinpoint and quantify scalability bottlenecks in fully optimized parallel programs with a measurement overhead of only a few percent. Recently, new capabilities were added to HPCTOOLKIT for collecting call path profiles for fully optimized codes without any compiler support, pinpointing and quantifying bottlenecks in multithreaded programs, exploring performance information and source code using a new user interface, and displaying hierarchical space-time diagrams based on traces of asynchronous call path samples. This paper provides an overview of HPCTOOLKIT and illustrates its utility for performance analysis of parallel applications.

99 GENERAL AND MISCELLANEOUS↗

Smart Contract Architectures and Templates for Blockchain-based Energy Markets (V.1.0)

Within the field of Transactive Energy Systems (TES), there is an active need for tools that can support and accelerate the development of these new grid solutions. Among the many tools available, blockchain stands out as a viable instrument that can help researchers develop decentralized, autonomous, and tamper-resistant grid applications. In this work, we explore the use of smart contracts (SCs), a subset of blockchain technology, and analyze their applicability to facilitating the implementation of TES solutions. In particular, we focus on presenting areas of opportunity and potential drawbacks, along with use cases that can benefit from this technology building upon previous research developed by Pacific Northwest National Laboratory and other research organizations. This work builds upon the fundamentals of TES and smart contract technology to develop a series of software templates that can be used by industry to build TES-oriented grid solutions. These templates are intended to be platform agnostic and take into consideration the unique properties of SCs and distributed ledger storage mechanisms to ensure actual code implementations remain aware of the limitations of the technology. The proposed templates have the potential to enable software architects to mix and match components to satisfy their application requirements, thereby reducing the number of resources required to implement blockchain-based solutions. These templates are divided into two main components—data and behavioral models. The data models are intended to help software engineers represent the underlying grid objects along with their properties in a ledger-based storage system. The behavioral models are used to describe the processes and actions that actors within a system must perform to achieve a given outcome such as registering an asset, placing a bid, and performing bid clearances. These two components are documented in a Unified Modeling Language (UML) format and are intended for use in SC-based implementations, with special behavioral considerations to account for the asynchronous properties of the underlying ledger and the typical execution model of smart contracts. Finally, future research ideas and potential extensions to this work are discussed. In particular, known limitations and potential improvements of the developed product are identified and expected to be addressed in future revisions of the template model.

24 POWER TRANSMISSION AND DISTRIBUTION↗

A burst compression and expansion technique for variable-rate users in satellite-switched TDMA networks

A burst compression and expansion technique is described for asynchronously interconnecting variable-data-rate users with cost-efficient ground terminals in a satellite-switched, time-division-multiple-access (SS/TDMA) network. Compression and expansion buffers in each ground terminal convert between lower rate, asynchronous, continuous-user data streams and higher-rate TDMA bursts synchronized with the satellite-switched timing. The technique described uses a first-in, first-out (FIFO) memory approach which enables the use of inexpensive clock sources by both the users and the ground terminals and obviates the need for elaborate user clock synchronization processes. A continous range of data rates from kilobits per second to that approaching the modulator burst rate (hundreds of megabits per second) can be accommodated. The technique was developed for use in the NASA Lewis Research Center System Integration, Test, and Evaluation (SITE) facility. Some key features of the technique have also been implemented in the gound terminals developed at NASA Lewis for use in on-orbit evaluation of the Advanced Communications Technology Satellite (ACTS) high burst rate (HBR) system.

Budinger, James M.↗

A burst compression and expansion technique for variable-rate users in satellite-switched TDMA networks

A burst compression and expansion technique is described for asynchronously interconnecting variable-data-rate users with cost-efficient ground terminals in a satellite-switched, time-division-multiple-access (SS/TDMA) network. Compression and expansion buffers in each ground terminal convert between lower rate, asynchronous, continuous-user data streams and higher-rate TDMA bursts synchronized with the satellite-switched timing. The technique described uses a first-in, first-out (FIFO) memory approach which enables the use of inexpensive clock sources by both the users and the ground terminals and obviates the need for elaborate user clock synchronization processes. A continuous range of data rates from kilobits per second to that approaching the modulator burst rate (hundreds of megabits per second) can be accommodated. The technique was developed for use in the NASA Lewis Research Center System Integration, Test, and Evaluation (SITE) facility. Some key features of the technique have also been implemented in the ground terminals developed at NASA Lewis for use in on-orbit evaluation of the Advanced Communications Technology Satellite (ACTS) high burst rate (HBR) system.

Budinger, James M.↗

Extending I/O through High Performance Data Services

The complexity of HPC systems has increased the burden on the developer as applications scale to hundreds of thousands of processing cores. Moreover, additional efforts are required to achieve acceptable I/O performance, where it is important how I/O is performed, which resources are used, and where I/O functionality is deployed. Specifically, by scheduling I/O data movement and by effectively placing operators affecting data volumes or information about the data, tremendous gains can be achieved both in the performance of simulation output and in the usability of output data. Previous studies have shown the value of using asynchronous I/O, of employing a staging area, and of performing select operations on data before it is written to disk. Leveraging such insights, this paper develops and experiments with higher level I/O abstractions, termed 'data services', that manage output data from 'source to sink': where/when it is captured, transported towards storage, and filtered or manipulated by service functions to improve its information content. Useful services include data reduction, data indexing, and those that manage how I/O is performed, i.e., the control aspects of data movement. Our data services implementation distinguishes control aspects - the control plane - from data movement - the data plane, so that both may be changed separably. This results in runtime flexibility not only in which services to employ, but also in where to deploy them and how they use I/O resources. The outcome is consistently high levels of I/O performance at large scale, without requiring application change.

97 MATHEMATICS AND COMPUTING↗

Fast GPU-Based Generation of Large Graph Networks From Degree Distributions

Synthetically generated, large graph networks serve as useful proxies to real-world networks for many graph-based applications. The ability to generate such networks helps overcome several limitations of real-world networks regarding their number, availability, and access. Here, we present the design, implementation, and performance study of a novel network generator that can produce very large graph networks conforming to any desired degree distribution. The generator is designed and implemented for efficient execution on modern graphics processing units (GPUs). Given an array of desired vertex degrees and number of vertices for each desired degree, our algorithm generates the edges of a random graph that satisfies the input degree distribution. Multiple runtime variants are implemented and tested: 1) a uniform static work assignment using a fixed thread launch scheme, 2) a load-balanced static work assignment also with fixed thread launch but with cost-aware task-to-thread mapping, and 3) a dynamic scheme with multiple GPU kernels asynchronously launched from the CPU. The generation is tested on a range of popular networks such as Twitter and Facebook, representing different scales and skews in degree distributions. Results show that, using our algorithm on a single modern GPU (NVIDIA Volta V100), it is possible to generate large-scale graph networks at rates exceeding 50 billion edges per second for a 69 billion-edge network. GPU profiling confirms high utilization and low branching divergence of our implementation from small to large network sizes. For networks with scattered distributions, we provide a coarsening method that further increases the GPU-based generation speed by up to a factor of 4 on tested input networks with over 45 billion edges.

97 MATHEMATICS AND COMPUTING↗

Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor

Due to the distributed and asynchronous nature of neural computation through low-energy spikes, brain-inspired hardware systems offer high energy efficiency and massive parallelism. One such platform is the IBM TrueNorth neurosynaptic system. Recently, TrueNorth compatible representation learning algorithms have emerged, achieving close to the state-of-the-art performance in various data sets. However, its application in temporal sequence processing models, such as recurrent neural networks (RNNs), is still only at the proof of concept level. There is an inherent difficulty in capturing temporal dynamics of an RNN using spiking neurons, which is only exasperated by the hardware constraints in connectivity and synaptic weight resolution. This paper presents a design flow that overcomes these difficulties and maps a special case of recurrent networks called long short-term memory (LSTM) onto a spike-based platform. The framework utilizes various approximation techniques, such as activation discretization, weight quantization, and scaling and rounding, spiking neural circuits that implement the complex gating mechanisms, and a store-and-release technique to enable neuron synchronization and faithful storage. While the presented techniques can be applied to map LSTM to any spiking neural network (SNN) simulator/emulator, here we choose the TrueNorth chip as the target platform by adhering to its hardware constraints. Three LSTM applications, parity check, extended Reber grammar, and question classification, are evaluated. The tradeoffs among accuracy, performance, and energy tradeoffs achieved on TrueNorth are demonstrated. Finally, this is compared with the performance on an SNN platform without hardware constraints, which represents the upper bound of the achievable accuracy.

97 MATHEMATICS AND COMPUTING↗

NASA Tech Briefs, October 2011

Topics covered include: Laser Truss Sensor for Segmented Telescope Phasing; Qualifications of Bonding Process of Temperature Sensors to Deep-Space Missions; Optical Sensors for Monitoring Gamma and Neutron Radiation; Compliant Tactile Sensors; Cytometer on a Chip; Measuring Input Thresholds on an Existing Board; Scanning and Defocusing Properties of Microstrip Reflectarray Antennas; Cable Tester Box; Programmable Oscillator; Fault-Tolerant, Radiation-Hard DSP; Sub-Shot Noise Power Source for Microelectronics; Asynchronous Message Service Reference Implementation; Zero-Copy Objects System; Delay and Disruption Tolerant Networking MACHETE Model; Contact Graph Routing; Parallel Eclipse Project Checkout; Technique for Configuring an Actively Cooled Thermal Shield in a Flight System; Use of Additives to Improve Performance of Methyl Butyrate-Based Lithium-Ion Electrolytes; Li-Ion Cells Employing Electrolytes with Methyl Propionate and Ethyl Butyrate Co-Solvents; Improved Devices for Collecting Sweat for Chemical Analysis; Tissue Photolithography; Method for Impeding Degradation of Porous Silicon Structures; External Cooling Coupled to Reduced Extremity Pressure Device; A Zero-Gravity Cup for Drinking Beverages in Microgravity; Co-Flow Hollow Cathode Technology; Programmable Aperture with MEMS Microshutter Arrays; Polished Panel Optical Receiver for Simultaneous RF/Optical Telemetry with Large DSN Antennas; Adaptive System Modeling for Spacecraft Simulation; Lidar-Based Navigation Algorithm for Safe Lunar Landing; Tracking Object Existence From an Autonomous Patrol Vehicle; Rad-Hard, Miniaturized, Scalable, High-Voltage Switching Module for Power Applications; and Architecture for a 1-GHz Digital RADAR.

Source record↗

Efficient I/O for Neural Network Training with Compressed Data

FanStore is a shared object store that enables efficient and scalable neural network training on supercomputers. By providing a global cache layer on node-local burst buffers using a compressed representation, it significantly enhances the processing capability of deep learning (DL) applications on existing hardware. In addition, FanStore allows POSIX-compliant file access to the compressed data in user space. We investigate the tradeoff between runtime overhead and data compression ratio using real-world datasets and applications, and propose a compressor selection algorithm to maximize storage capacity given performance constraints. We consider both asynchronous (i.e., with prefetching) and synchronous I/O strategies, and propose mechanisms for selecting compressors for both approaches. Using FanStore, the same storage hardware can host 2-13x more data for example applications without significant runtime overhead. Empirically, our experiments show that FanStore scales to 512 compute nodes with near linear performance scalability.

Zhang, Zhao↗

DARMA-MPI Interoperability WBS 2.3.1.04 Milestone 15 (ECP Milestone Report)

DARMA (Distributed Asynchronous Resilient Models for Applications) is a runtime library developed as part of the the Sandia ATDM (Advanced Technology Development and Mitigation) program. DARMA supports applications within 2.2.5.03 ADNN03-ASC ATDM SNL Application, which includes a number of applications featuring dynamic physics which requires load balancing and asynchronous communication for high performance. We have implemented a modern C++ programming model that can enable dynamic, asynchronous communication on top of existing data structures from a serial or MPI code. DARMA development has occurred in parallel with a verification milestone for ATDM in FY18. For FY19, DARMA will impact ATDM by enabling the performance benefits of a dynamic runtime through only incremental changes to existing verified MPI codes. The results presented here demonstrate the DARMA development process for an MPI mini-app, showing a 3-4x improvement in performance for a challenging problem relative to the parent MPI code and coming within 25 percent of the theoretically optimal performance achievable from a perfect, fine-grained load balancer for most cases.

97 MATHEMATICS AND COMPUTING↗

Collecting Real-Time Test Data

The TAXI Direct-to-Direct interface, in conjunction with the Disk Recording System (DRS) software developed by the SSC Boeing group, supplies a solution that provides higher integrity data that can be processed and made available to the customer with hours. The TAXI Direct-to-Direct interface is a special purpose interface solution developed by ISC for demultiplexing data from a Racal Storeplex (or equivalent) multichannel recorder onto one or more hard discs that reside in, and/or are controlled by, a personal computer (PC). (The name "TAXI" is the acronym for "transparent asynchronous transceiver interface"). The installation of the TAXI Direct-to Disk interface, in conjunction with other modifications, causes the transfer of data to be recorded directly to disk so it can be extracted and analyzed after the test.

Source record↗

Spatial and Temporal Organization of Chromosome Duplication and Segregation in the Cyanobacterium Synechococcus elongatus PCC 7942

The spatial and temporal control of chromosome duplication and segregation is crucial for proper cell division. While this process is well studied in eukaryotic and some prokaryotic organisms, relatively little is known about it in prokaryotic polyploids such as Synechococcus elongatus PCC 7942, which is known to possess one to eight copies of its single chromosome. Using a fluorescent repressor-operator system, S. elongatus chromosomes and chromosome replication forks were tagged and visualized. We found that chromosomal duplication is asynchronous and that the total number of chromosomes is correlated with cell length. Thus, replication is independent of cell cycle and coupled to cell growth. Replication events occur in a spatially random fashion. However, once assembled, replisomes move in a constrained manner. On the other hand, we found that segregation displays a striking spatial organization in some cells. Chromosomes transiently align along the major axis of the cell and timing of alignment was correlated to cell division. This mechanism likely contributes to the non-random segregation of chromosome copies to daughter cells.

59 BASIC BIOLOGICAL SCIENCES↗

Alternative majority-voting methods for real-time computing systems

Two techniques that provide a compromise between the high time overhead in maintaining synchronous voting and the difficulty of combining results in asynchronous voting are proposed. These techniques are specifically suited for real-time applications with a single-source/single-sink structure that need instantaneous error masking. They provide a compromise between a tightly synchronized system in which the synchronization overhead can be quite high, and an asynchronous system which lacks suitable algorithms for combining the output data. Both quorum-majority voting (QMV) and compare-majority voting (CMV) are most applicable to distributed real-time systems with single-source/single-sink tasks. All real-time systems eventually have to resolve their outputs into a single action at some stage. The development of the advanced information processing system (AIPS) and other similar systems serve to emphasize the importance of these techniques. Time bounds suggest that it is possible to reduce the overhead for quorum-majority voting to below that for synchronous voting. All the bounds assume that the computation phase is nonpreemptive and that there is no multitasking.

Shin, Kang G.↗

Fit2Fly: A Proof of Concept for Testing the Commercial Feasibility of Unmanned Aerial System Operations

As autonomous Unmanned Aerial Systems (UAS) become more prevalent, improvements in airworthiness assessments, data security, establishing vehicle trust, and more will become necessary to make fleet operations of these systems routine. Currently, these are verified for every UAS vehicle through a time-intensive manual inspection performed by human individuals. This process becomes infeasible when introduced into large-scale operations, so the Fit2Fly project is attempting to address these concerns. Fit2Fly is a multi-year project with a large scope, so we chose to narrow our focus to building a proof-of-concept feasibility simulation in tandem with physical UAS demonstration. This work resulted in a software simulation capable of running asynchronous commercial operations with persistent data modeling in coordination with several autonomous drones using integrated radio and GPS configurations.

Ryan Bonk↗