民航气象数值预报系统高性能集群平台性能测试分析

利用高性能集群计算机进行气象数值预报计算,是提高预报准确率的有效途径,而对高性能集群的计算能力进行客观评估,是掌握设备性能、提高计算能力的基础,同时也对硬件选型、项目规划具有极大的参考价值。在现有的民航气象数值预报平台上,利用国际上最为流行的基准测试算法LINPACK,对平台的浮点计算能力进行了测试,评估了现有系统的性能,为进一步提高集群性能奠定基础。     

基于WSRF的集合预报任务调度系统

分析集合预报的任务特点,基于Web服务资源框架(WSRF)设计一种动态自适应的集合预报任务调度系统。给出任务调度算法,根据网格节点的资源变化自适应地选择节点,为每对初值样本的模式预报任务指派CPU资源数目,以缩短预报时间、节省计算成本。实际运行结果验证了该系统的可行性。     

基于抢占阈值调度的周期任务请求的响应时间计算

在实时系统中计算任务请求的响应时间具有重要意义,针对抢占阈值调度,给出了一个计算任务请求响应时间的公式。试验结果表明所给公式与采用模拟运行的方法相比具有计算量小、运行时间短的优点。     

基于知识图谱的海洋数值预报数据推荐算法

为解决海洋数值预报研究人员面对复杂多样的研究任务时难以及时准确地从种类繁多的海洋数值预报数据中找到所需数据的问题，提出基于知识图谱的海洋数值预报数据推荐算法。利用海洋数值预报文献提取研究任务及海洋数值预报数据构建知识图谱，基于知识图谱计算海洋数值预报数据实体之间的相似度，同时融合在研究人员用户行为下海洋数值预报数据的相似度，进行排序选取相似度较高的海洋数值预报数据进行推荐。实验结果表明，推荐精确率及召回率分别为67.14%、62.49%。     

利用GraphChi大数据技术解决相似预报中的图谱计算

图谱计算是相似预报中解决天气现象关联关系、不同时刻相似匹配的核心问题.其复杂的搜索分析和拟合比对任务产生巨大工作量,成为发展相似预报的计算瓶颈.利用GraphChi大数据技术,将大量集中式工作分散到多台电脑上完成,用海量硬盘替代集中式内存的访存瓶颈,为相似预报中图谱计算提供了新的解决方案.     

基于Spark+YARN的空气污染预报云平台设计

近年来我国空气污染导致的雾霾天气频繁发生,空气污染已经成为一个亟待解决的重要问题,其数值预报推广应用是一个难题,主要表现在数值预报系统的运行过程中计算量大,数据量大,在传统高性能计算集群中进行空气污染预报存在资源有限、复杂的并行化操作、批处理作业等待耗时等问题,尤其对于资源缺乏的研究团队,还存在基础设施昂贵而无法负担的问题。因此,如何利用有限的资源,为大气科学家提供一种基于云计算模式的高性能计算环境,提供可扩展、快速、廉价和动态可分配的计算和存储资源,是亟待解决的关键问题。研究一种基于Spark+YARN的空气污染数值预报的云平台,针对空气污染数值预报特点,为大气科学家提供一种空气污染数值预报大数据解决方案。     

基于模糊聚类神经网络的化工过程预报模型研究

本文提出了在化工过程预报中的模糊聚类神经网络模型,该模型具有提取典型数据、优化模糊规则及优化参数的优点,在化工过程预报实验中与传统方法相比预报结果的精度提高,计算时间缩短。     

OPTICS算法在雷电临近预报中的应用

针对密度分布不均的雷电定位资料,提出了一种基于OPTICS聚类算法的雷电临近预警模型。该模型运用OPTICS算法对雷暴天气连续时段的雷电定位资料进行聚类分析,有效剔除了影响雷暴云分布的稀疏点。在聚类分析结果基础上,利用"膨胀-侵蚀"算法还原雷暴云真实分布,根据雷暴云的移动趋势进行雷电落区预报。此外,针对传统预测算法运行时间长的缺陷,运用邻接表改进了OPTICS算法,且优化了可达队列更新策略。实验结果表明,基于改进的OPTICS算法所构建的雷电临近预报模型降低了算法运行时间,同时提高了雷电预报模型适应能力及预测的准确率。     

基于流水线模式的河系径流预报并发计算研究

在河系径流预报计算中,一方面受单站水文过程计算复杂性影响,另一方面下游站点依赖上游关联节点,现有洪水预报系统在河系预报计算时多采用串联模式进行计算。这在河流系预报节点较多、模型方法略为复杂时,计算效率较低。为突破河系径流预报计算效率瓶颈,本研究引入流水线并行模式,对河系径流预报站点初始化、单元产汇流计算、河道洪水演算、校正分析等模块进行拆解,构建流水线式工作站,将径流预报站点按水力联系连续入站,实现河系节点集径流过程的平行并发计算。选取淮河正阳关以上流域50余断面进行了模拟试验,结果表明：研究构建的并发计算方法计算结果可靠,较串行结构效率提升超3倍,可满足洪水预报实时性要求、尤其适用于B/S模式对系统响应效率的需求。     

无线局域网业务流预报方法比较

介绍了基于时间序列、神经网络和小波的多种网络业务的预报方法,应用真实的无线局域网业务流序列检验了这些模型的预报性能,结果表明,和其他预报模型相比,基于神经网络的模型能够比较精确地捕获无线局域网业务流自身的特性,对业务流具有良好的预报性能,而基于ARIMA模型的预报性能最差。     

Performance implications of synchronization structure in parallel programming

The restricted synchronization structure of so-called structured parallel programming paradigms has an advantageous effect on programmer productivity, cost modeling, and scheduling complexity. However, imposing these restrictions can lead to a loss of parallelism, compared to using a programming approach that does not impose synchronization structure. In this paper we study the potential loss of parallelism when expressing parallel computations into a programming model which limits the computation graph (DAG) to series–parallel topology, which characterizes all well-known structured programming models. We present an analytical model that approximately captures this loss of parallelism in terms of simple parameters that are related to DAG topology and workload distribution. We validate the model using a wide range of synthetic and real-world parallel computations running on shared and distributed-memory machines. Although the loss of parallelism is theoretically unbounded, our measurements show that for all above applications the performance loss due to choosing a series–parallel structured model is invariably limited up to 10%. In all cases, the loss of parallelism is predictable provided the topology and workload variability of the DAG are known.     

基于多输出学习的沪深300指数预测研究

在股票市场中,人们通常会依赖于股票的历史交易数据来进行推测。目前采用SVM方法进行预测的研究较多,但其模型复杂,耗时较长,而且通常只预测未来1天的数据。文中采用多输出的正则化方法来预测未来多天的走势,对多任务的学习方法进行改进,提出了一种基于多输出的学习方法。实验表明,与SVM支持向量机的方法相比,该方法在沪深300指数预测的均方差值上提高了约10倍,运行时长也减少了近3/4。     

Predict task running time in grid environments based on CPU load predictions

A good running time prediction of tasks is very helpful and important for job scheduling and resource management in grid systems. In this paper, we present a running time prediction method for grid tasks based on our previous work, which is a novel CPU load prediction method. In order to eliminate the interference of other factors, such as memory accessing, network performance, and fluctuation of competing CPU load and so on, we produce a simulation to test and evaluate our prediction method. In this simulation we use more than 10,000 randomized test cases run on load traces sampled from 39 different machines. The simulation results are excellent and demonstrate that our running time prediction of grid tasks outperforms significantly that of a widely existing prediction method.     

静态并行任务调度模拟工具的设计与实现

介绍构建的一个包含12种静态并行任务调度算法的模拟器TSAS和一个可视化的任务图生成器VTGG,并给出以它们为工具得到的各种任务调度算法的性能数据.     

Techniques of subjective workload assessment: a comparison of SWAT and the NASA-Bipolar methods

Abstract

Assessment of subjective workload is becoming increasingly important in the evaluation of human-machine systems. Two popular methods were compared: (1) the Subjective Workload Assessment Technique (SWAT) that employed a conjoint measurement procedure to confer interval scale properties on the workload ratings, and (2) a technique under development at NASA that used an individually weighted workload score. Both methods were applied in a laboratory experiment that required the rating of a number of single- and dual-tracking and spatial transformation tasks. Both subjective assessment techniques displayed similar sensitivity to the different task manipulations. However, both techniques failed to detect the resource competition effects in the dual-task performance, and were in general insensitive to response execution processing demands. A notable difference between the two techniques was that the NASA-Bipolar ratings consistently had a smaller between-subject variability than the SWAT ratings. Discussion of the results is centred around the issue of the validity of assessment of subjective workload in general, and the construct and concurrent validity of the two techniques in particular.     

ScopeWork中任务处理机制的设计

工作流管理系统（ＷＦＭＳ）把业务流程中的每一项活动着作任务,并负责组织、调度和控这些相互关联的任务。工作流中任务的模型描述和接口设计是开发工作流应用中重要的组成部分,它直接影响着任务间的调度执行,本文主要讨论基于ＣＯＲＢＡ的ＷＦＭＳ ＳｃｏｐＷｏｒｋ中的任务管理器和任务处理的设计。     

卫星应用任务分解规则库建立及匹配算法

卫星应用任务的分解是准确分析航天信息系统使命任务的关键步骤,为此需要对任务的分解建立标准的规则库.文章基于Disciple代理的知识表达方法,建立了卫星应用任务的分解规则库;然后通过一般化的概念对任务和规则进行分析,提出了一种任务匹配规则库的算法.文章以卫星应用中典型的侦察任务为例,探讨了规则库的建立和匹配.可以看出Disciple灵活的表达可以有效地描述好所涉及对象及其之间的关系,满足规则库建立的需要.     

Modelling complex processes in GTA

GTA is a method for GroupWare task analysis. Representations of task models in GTA originally focused on object oriented templates of the various elements, and on hierarchical relations between elements as seen from separate viewpoints (work, agents and situation). In order to model time aspects and complex processes, the authors introduced new concepts as well as new types of representation inspired by workflow representation techniques.     

JAPS中任务调度协议的设计与实现

1 背景和系统概述在并行编译系统中,主要的目标是分析程序中的依赖关系。通过依赖分析得到的结果将源程序划分成可以并行执行的部分,然后在一定的运行时支撑平台上并行执行。并行执行的基本要求是程序的并行语义和串行语义必须一致。在以NOW基础的并行计算系统中,由于通信开销较大,一般需要发掘在任务层次上的并行性。相对数     

Keeping found things found: the study and practice of personal information management

Abstract

The application of task analysis to assess the adequacy of a proposed design for an independent feed system for use on a number of nuclear power stations is described. The objectives of the analysis are identified and the methodology to achieve these objectives is developed. The analysis examines the demands made upon station operations staff to commission the proposed feed system. The impact of potential operator error and adverse environmental conditions on commissioning time is considered. The results of the analyses and their input into the design process are summarized. A subsequent comparison between the results of the analysis and a demonstration on an installed system is described and differences discussed. Finally the relative merits of computer-based task simulation techniques in this type of exercise are considered.