Search NASA⌕ Search

SEARCH · Search NASA

Results for “scale-out”

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.

FatMan vs. LittleBoy: Scaling Up Linear Algebraic Operations in Scale-Out Data Platforms

Linear algebraic operations such as matrix manipulations form the kernel of many machine learning and other crucial algorithms. Scaling up as well as scaling out such algorithms are highly desirable to enable efficient processing over millions of data points. To this end, we present a matrix manipulation approach to effectively scale-up each node in a scale-out data parallel platform such as Apache Spark. Specifically, we enable hardware acceleration for matrix multiplications in a distributed Spark setup without user intervention. Our approach supports both dense and sparse distributed matrices, and provides flexible control of acceleration by matrix density. We demonstrate the benefit of our approach for generalized matrix multiplication operations over large matrices with up to four billion elements. To connect the effectiveness of our approach with machine learning applications, we performed Gramian matrix computation via generalized matrix multiplications. Our experiments show that our approach achieves more than 2× performance speed-up, and up to 96.1% computation improvement, compared to a state of the art Spark MLlib for dense matrices.

97 MATHEMATICS AND COMPUTING↗

BESPOKV: Application Tailored Scale-Out Key-Value Stores

Enterprise KV stores are not well suited for HPC applications, and entail customization and cumbersome end-to-end KV design to extract the HPC application needs. To this end, in this paper we present BESPOKV, an adaptive, extensible, and scale-out KV store framework. BESPOKV decouples the KV store design into the control plane for distributed management and the data plane for local data store. BESPOKV takes as input a single-server KV store, called a datalet, and transparently enables a scalable and fault-tolerant distributed KV store service. The resulting distributed stores are also adaptive to consistency or topology requirement changes and can be easily extended for new types of services. Experiments show that BESPOKV-enabled distributed KV stores scale horizontally to a large number of nodes, and performs comparably and sometimes better than the state-of-the-art systems.

Anwar, Ali↗

Customizable Scale-Out Key-Value Stores

Enterprise KV stores are often not well suited for HPC applications, and thus cumbersome end-to-end KV design customization is required to meet the needs of modern HPC applications. To this end, in this article we present bespoKV, an adaptive, extensible, and scale-out KV store framework. bespoKV decouples the KV store design into the control plane for distributed management and the data plane for local data store. For the control plane, bespoKVprovides pre-built modules, called controlets, supporting common distributed functionalities (e.g., replication, consistency, and topology) and their various combinations. This decoupling allows bespoKV to take a user-provided single-server KV store, called a datalet, and transparently enables a scalable and fault-tolerant distributed KV store service. The resulting distributed stores are also adaptive to consistency or topology requirement changes and can be easily extended for new types of services. Such specializations enable innovative uses of KV stores in HPC applications, especially for emerging applications that utilize KV-friendly workloads. We evaluate bespoKV in a local testbed as well as in a public cloud settings. Experiments show that bespoKV-enabled distributed KV stores scale horizontally to a large number of nodes, and performs comparably and sometimes 1.2× to 2.6× better than the state-of-the-art systems.

97 MATHEMATICS AND COMPUTING↗

FatMan vs. LittleBoy: Scaling up Linear Algebraic Operations in Scale-out Data Platforms

Linear algebraic operations such as matrix manipulations form the kernel of many machine learning and other crucial algorithms. Scaling up as well as scaling out such algorithms are highly desirable to enable efficient processing over millions of data points. To this end, we present a matrix manipulation approach to effectively scale-up each node in a scale-out data parallel platform such as Apache Spark. Specifically, we enable hardware acceleration for matrix multiplications in a distributed Spark setup without user intervention. Our approach supports both dense and sparse distributed matrices, and provides flexible control of acceleration by matrix density. We demonstrate the benefit of our approach for generalized matrix multiplication operations over large matrices with up to four billion elements. To connect the effectiveness of our approach with machine learning applications, we performed Gramian matrix computation via generalized matrix multiplications. Our experiments show that our approach achieves more than 2x performance speed-up, and up to 96.1% computation improvement, compared to a state of the art Spark MLlib for dense matrices.

Lim, Seung-Hwan↗

Tartan: Evaluating Modern GPU Interconnect via a Multi-GPU Benchmark Suite

In this paper, we fill the gap by proposing a multi-GPU benchmark suite named Tartan, which contains microbenchmarks, scale-up and scale-out applications. We then apply Tartan to evaluate the four latest types of modern GPU interconnects, i.e., PCI- e, NVLink-V1, NVLink-V2 and InfiniBand with GPUDirect- RDMA from two recently released NVIDIA super AI platforms as well as ORNL’s exascale prototype system. Based on empirical evaluation, we observe four new types of NUMA effects: three types are triggered by NVLink’s topology, connectivity and routing, while one type is caused by PCI-e (i.e., anti-locality). They are very important for performance tuning in multi-GPU environment. Our evaluation results show that, unless the current CPU-GPU master-slave programming model can be replaced, it is difficult for scale-up multi-GPU applications to really benefit from faster intra-node interconnects such as NVLinks; while for inter-node scale-out applications, although interconnect is more crucial to the overall performance, GPUDirect-RDMA appears to be not always the optimal choice.

HPC↗

T-FSM: A Scalable Distributed Task-Based System for Frequent Subgraph Pattern Mining from a Big Graph

Finding frequent subgraph patterns in a big graph is an important problem with many applications such as classifying chemical compounds and building indexes to speed up graph queries. Since this problem is NP-hard, some recent parallel and distributed systems have been developed to accelerate the mining. However, they often have a huge memory cost, very long running time, suboptimal load balancing, poor scale-out capability, and possibly inaccurate results. In this article, we propose an efficient system called T-FSM for parallel mining of frequent subgraph patterns in a big graph. T-FSM supports a new anti-monotonic frequentness measure called Fraction-Score, which is more accurate than the widely used MNI measure. The execution engine of T-FSM supports both intra-machine parallelism and inter-machine parallelism. For intra-machine parallelism, T-FSM adopts a novel task-based execution model to ensure high multithreading concurrency, bounded memory consumption, and effective load balancing. For inter-machine parallelism, T-FSM ensures good scale-out performance with a lightweight pattern rebalancing approach that reduces workload skewness of pattern evaluations among machines. To avoid recomputing the contexts for migrated patterns, we design a novel context cache table to support concurrent and asynchronous requesting and caching of remote context data, which can timely evict and garbage collect used pattern contexts that are no longer needed to keep memory consumption bounded. Extensive experiments show that T-FSM is orders of magnitude faster than existing state-of-the-art parallel systems (more than 10×, 51×, 131×, 55× speedup over ScaleMine, DistGraph, Pangolin and Peregrine, respectively) and distributed systems (more than 42× and 88× over ScaleMine and DistGraph, respectively) for frequent subgraph pattern mining, and it scales out satisfactorily to 512 CPU cores on the Polaris supercomputer at Argonne National Laboratory.

97 MATHEMATICS AND COMPUTING↗

Hydrogen Production System Scaling Using a High-Fidelity Simulation-Optimization Framework

Proton exchange membrane (PEM) electrolyzers are widely used for hydrogen production, yet few validated, high-fidelity tools can reliably guide scale-up. Using measured performance from a 50-hour hardware-in-the-loop pilot test, a physics-based, plant-level model of a 1.25 MW PEM electrolyzer and its balance-of-plant (BoP) subsystems is developed and validated. The model couples electrochemistry and thermal/flow submodels and is calibrated against pilot test data via a genetic algorithm (GA) workflow. Validation yields a mean absolute percentage error (APE) of 0.43% for cell voltage and stack power. Two scale-out strategies are then benchmarked under a common 7-day wind-and-photovoltaic (PV) profile: (i) linear duplication of 1.25 MW blocks and (ii) shared-BoP architectures. Sharing BoP between stacks reduces BoP energy by 27% at 10 MW and 34% at 100 MW (vs. linear duplication) and improves system specific energy consumption (SEC) to 52.9 and 52.6 kWh/kg, respectively (from 54.0 kWh/kg with linear duplication). Partial-load studies (25-100% set-point) show that cumulative hydrogen production remains nearly constant down to 50% load because all cases use the same weekly renewable-energy input. Below 50%, the power cap limits how much energy can be used within 168 h, which reduces hydrogen output. The model further indicates that the practical operating optimum lies between 50% and 85% load, where efficiency gains begin to appear without significant loss in hydrogen output. Moreover, the efficiency gains at lower loads are offset by reduced production. The validated framework supports scenario-based engineering trade-off studies for large configurations (10-100 MW) and for operating policies under variable renewables.

08 HYDROGEN↗

X-ray transparent microfluidic chips for high-throughput screening and optimization of in meso membrane protein crystallization

Elucidating and clarifying the function of membrane proteins ultimately requires atomic resolution structures as determined most commonly by X-ray crystallography. Many high impact membrane protein structures have resulted from advanced techniques such as in meso crystallization that present technical difficulties for the set-up and scale-out of high-throughput crystallization experiments. In prior work, we designed a novel, low-throughput X-ray transparent microfluidic device that automated the mixing of protein and lipid by diffusion for in meso crystallization trials. Here, in this work, we report X-ray transparent microfluidic devices for high-throughput crystallization screening and optimization that overcome the limitations of scale and demonstrate their application to the crystallization of several membrane proteins. Two complementary chips are presented: (1) a high-throughput screening chip to test 192 crystallization conditions in parallel using as little as 8 nl of membrane protein per well and (2) a crystallization optimization chip to rapidly optimize preliminary crystallization hits through fine-gradient re-screening. We screened three membrane proteins for new in meso crystallization conditions, identifying several preliminary hits that we tested for X-ray diffraction quality. Further, we identified and optimized the crystallization condition for a photosynthetic reaction center mutant and solved its structure to a resolution of 3.5 Å.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Breaking the mold: Overcoming the time constraints of molecular dynamics on general-purpose hardware

The evolution of molecular dynamics (MD) simulations has been intimately linked to that of computing hardware. For decades following the creation of MD, simulations have improved with computing power along the three principal dimensions of accuracy, atom count (spatial scale), and duration (temporal scale). Since the mid-2000s, computer platforms have, however, failed to provide strong scaling for MD, as scale-out central processing unit (CPU) and graphics processing unit (GPU) platforms that provide substantial increases to spatial scale do not lead to proportional increases in temporal scale. Important scientific problems therefore remained inaccessible to direct simulation, prompting the development of increasingly sophisticated algorithms that present significant complexity, accuracy, and efficiency challenges. While bespoke MD-only hardware solutions have provided a path to longer timescales for specific physical systems, their impact on the broader community has been mitigated by their limited adaptability to new methods and potentials. In this work, we show that a novel computing architecture, the Cerebras wafer scale engine, completely alters the scaling path by delivering unprecedentedly high simulation rates up to 1.144 M steps/s for 200 000 atoms whose interactions are described by an embedded atom method potential. This enables direct simulations of the evolution of materials using general-purpose programmable hardware over millisecond timescales, dramatically increasing the space of direct MD simulations that can be carried out. In this paper, we provide an overview of advances in MD over the last 60 years and present our recent result in the context of historical MD performance trends.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Exotanium DOE SBIR Phase I Results Summary

Exotanium demonstrated this technology with the Idaho National Laboratory’s MASTODON application, a Multiphysics environment designed to run typical high-performance computing (HPC) simulations for structural dynamics, seismic analysis, and risk assessment. The MASTODON application was packaged into a container using Docker, Deployed on Amazon ECS, and managed through a custom Scale-Out Compute on AWS (SOCA) implementation.

97 MATHEMATICS AND COMPUTING↗

Exploring DAOS as a Burst Buffer for a 100 Gbps DAQ Real-Time Streaming System

We present an experimental evaluation of a burst buffer for a real-time DAQ streaming system designed to transmit instrument data to remote data centers. The system is based on EJ-FAT, a load balancing system capable of Nx 100Gbps streams, distributing data from event sources to processing nodes. We explore applying the DAOS system as a burst buffer to serve a number of purposes: improve resiliency, elasticity and add new functions into the processing pipeline. In the evaluation a sender transmits events over a 100Gbps network to a receiver integrated with DAOS to store the reassembled events using DAOS APIs. We evaluate the system for possible bottlenecks and provide end-to-end evaluation with a burst buffer using DAOS storage abstractions. We show that a receiver node can support 38.1 Gbps. This proves the viability of our approach and allows us to extend this work to investigate scale-out properties and new streaming optimizations.

Mei, Xinxin↗

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↗

Data Movement Categories

We have endeavored to classify the commonly seen data movement needs, as observed in data-intensive institutions (both commercial and non-profit), into four categories. Knowing how to map a data movement task into one of the four categories helps select proper data mover tools. For each category, how the data storage is involved, high-level examples and the nature of typical solutions are described. Finally, some general remarks are provided to help further orient readers new to this field - the 4th IT pillar.

99 GENERAL AND MISCELLANEOUS↗

When to use rsync

We have endeavored to show, using a series of data transfer results obtained from two testbeds, when to use the popular data copying tool rsync and related tools. Tests have been conducted in local area network (LAN) and wide area network (WAN) environments. We conclude that for files in a certain size range and network latency ≦ 10 ms round trip time (RTT), rsync is still useful for data moving tasks in the category 4 of the U.S. DOE Technical Report “Data Movement Categories”. For more demanding data movement requirements, tools of different classes are suggested. Sample histograms from two DOE user facilities are provided to further support our conclusions.

97 MATHEMATICS AND COMPUTING↗