基于SOAP的多机群网格通信机制的研究与实现

为了在由多计算机机群构成的多机群网格环境下利用SOAP协议进行并行计算,提出了一个采用SOAP的多机群网格并行计算通信模型。根据Adhoc的自组织机制、容错机制和压缩机制实现了该模型。通过SOAP与TCP传输性能对比的结果表明,该模型可以有效支持多机群网格的构建及其上的并行通信。     

基于SOAP的多机群网格通信机制的研究与实劲

为了在由多计算机机群构成的多机群网格环境下利用SOAP协议进行并行计算，提出了一个采用SOAP的多机群网格并行计算通信模型。根据Adhoe的自组织机制、容错机制和压缩机制实现了该模型。通过SOAP与TCP传输性能对比的结果表明，该模型可以有效支持多机群网格的构建及其上的并行通信。     

基于二维滑动窗口的数据网格通信模型

在由多计算机集群构成的数据密集型网格环境下,为了解决通信峰值阶段的通信冲突问题,提出了一个基于二维滑动窗口机制的通信模型。给出了由多计算机机群构成的数据密集型网格数据节点DGDN(data-intensivegriddatanode)、完全图通信和二维滑动窗口等一系列形式化定义;利用计算节点和网络的通信能力,研究了二维滑动窗口机制,描述了单个机群内的完全图通信算法CA2DSW(communicationalgorithmbased2D-sliderwindow)以及双机群内的计算机之间的完全图通信算法CADC(communicationalgorithmindoublecluster)。基于网络距离机制、CA2DSW和CADC,设计了DGDN上的通信模型CMDGDN。理论分析和实验结果表明,该模型有效地解决了数据网格环境下通信峰值所造成的并行计算效率低下的问题。该模型适合于基于网格的数据密集型并行计算。     

基于任务映射的并行辐射度算法

基于辐射度算法的大规模数据场景的可视化模拟中所面临的主要问题是其计算时间无法满足实时性的要求,提出了任务映射模型较好地解决了各计算节点的负载均衡性要求,从而有效提高辐射度算法的并行计算效率。基于该模型开发了并行辐射度算法并在超级计算机模拟器上进行了测试,试验结果表明,该算法具有良好的可扩展性和较高的计算加速比。     

校园计算网格作业自适应调度的研究和实现

由于网格资源具有分布、异构、动态、自治等特性。使得网格作业调度成为网格计算中具有挑战性的任务之一。针对校园高性能计算资源负载不均衡、管理模式不统一的问题，提出了校园计算网格自适应调度系统的解决方案。详细研究了网格作业的白适应调度技术，设计了含有负载均衡能力的资源动态信息收集模型，并给出自适应调度算法。实现了校园范围内高性能计算资源的负载均衡和统一的资源管理。     

JOGR——利用Java对象相关系支持多模式并行编程

为支持在机群环境和Grid环境中实现Master/Slave、“分而治之”、“阶段并行”等多种并行范例，本文提出了Master/Group/Slave编程模式JOGR可以实现机群环境下的动态数据分配和负载均衡。此外，我们还基于WebServic技术和Jave语言实现了JOGR。实验结果表明，JOGR能为并行程序的开发运行提供有效的支持。     

JOGR--利用Java对象组关系支持多模式并行编程

为支持在机群环境和Grid环境中实现Master/Slave、"分而治之"、"阶段并行"等多种并行范例,本文提出了Master/Group/Slave编程模式JOGR,利用JOGR可以实现机群环境下的动态数据分配和负载均衡.此外,我们还基于WebServic技术和Java语言实现了JOGR.实验结果表明JOGR能为并行程序的开发运行提供有效的支持.     

基于多Agent网格资源管理模型的负载均衡研究

网格是下一代互联网的应用模式，资源管理是网格技术的核心任务之一。它包含资源发现、任务调度和负载均衡。在网格环境中，当大量任务到达的时候，总会出现某些节点因任务过多而忙，另外一些节点空闲的情况。为了解决这个问题．文章首先介绍了一种基于多Agent的网格资源管理层次模型，在此基础上研究了负载均衡问题，提出了一种负载均衡的算法设计思想。提高了网格计算能力和资源利用率。     

地球物理并行计算平台的研究

随着地球物理数据处理计算量的快速增加,使得单机运算已难以满足其计算需求。采用并行计算技术来提高处理速度是主要发展方向。为了解决地球物理数据处理中的大规模计算问题,提出基于工作站集群的并行系统架构,采用JPPF的网格计算框架,构建地球物理并行计算平台。JPPF易于使用、代码开源,利用其在并行任务管理和负载均衡方面的优势,设计并实现了一个2.5维可控源音频大地电磁正演并行系统。在实验中对频率并行和非并行的计算效率进行了对比。实验结果表明在并行计算平台硬件一定的情况下,随着频点个数的增多,该系统能够较为明显地提高计算效率。     

高效训练百万亿参数预训练模型的系统挑战和对策

随着预训练模型规模的急剧增长,训练此类模型需要海量的计算和存储能力。为此,本工作在新一代国产高性能计算机上训练了一个174万亿参数的超大规模预训练模型,模型参数量可与人脑中的突触数量相媲美。重点讨论在训练这一超大规模预训练模型中遇到的几个关键系统挑战：如何选取高效并行策略,如何进行高效数据存储,如何选取合适的数据精度,以及如何实现动态负载均衡,并总结了针对上述挑战的一些解决方法。     

A parallel implementation of NEC for the analysis of large structures

We present a new, parallel version of the numerical electromagnetics code (NEC). The parallelization is based on a bidimensional block-cyclic distribution of matrices on a rectangular processor grid, assuring a theoretically optimal load balance among the processors. The code is portable to any platform supporting message passing parallel environments such as message passing interface and parallel virtual machine, where it could even be executed on heterogeneous clusters of computers running on different operating systems. The developed parallel NEC was successfully implemented on two parallel supercomputers featuring different architectures to test portability. Large structures containing up to 24000 segments, which exceeds currently available computer resources were successfully executed and timing and memory results are presented. The code is applied to analyze the penetration of electromagnetic fields inside a vehicle. The computed results are validated using other numerical methods and experimental data obtained using a simplified model of a vehicle (consisting essentially of the body shell) illuminated by an electromagnetic pulse (EMP) simulator.     

一类大规模油藏数值模拟问题的有效并行计算

大型科学计算领域的并行处理一直是人们非常关注,如油藏数值模拟的并行计算,本文基于用网络连接多台共离内存多处理机的分布并行处理系统,运用并行计算理论和方法,采用负载衡策略,解决了网络节点多达７２．８万的大规模油藏数值模拟问题,提高了油藏数值模拟的计算速度和能力。     

基于Web Service技术分布式并行数据挖掘的研究与实现

主要介绍基于Web Service技术的一个数据挖掘系统,在一个关联规则挖掘的并行算法—CD算法的基础上,结合一种基于动态数据集划分的并行关联规则挖掘算法,利用动态方式分配数据量,使每个处理器获得相同多的数据集,解决在网络中大量分散的数据因通信等问题而引起的负载平衡,从而提高了数据挖掘效率。     

三维电磁散射的网络并行FDTD计算和加速比分析

应用计算机局域网,采用基于消息传递PVM平台和区域分解技术,实现了三维电磁散射的并行FDTD计算.给出了在FDTD两个相邻子区域交界面上所需要传递数据量的估算和分析.在一个实际的计算机局域网环境下,测试了网络并行FDTD计算三维机翼目标散射时的并行加速比和并行效率.实际上,并行加速比和效率不仅与局域网的硬件性能有关,而且与子区域的划分和PVM通信原语的使用等软件设计有关.最后,讨论并行计算中的附加通信量、网络通信性能和负载平衡对FDTD并行计算的影响.     

Grid‐Enabled Parallel Simulation Based on Parallel Equation Formulation

Parallel simulation is an efficient way to cope with long runtimes and high computational requirements in simulations of modern complex integrated electronic circuits and systems. This paper presents an algorithm for parallel simulation based on parallelization in equation formulation and simultaneous calculation of matrix contributions for nonlinear analog elements. In addition, the paper describes the development of a grid interface for a parallel simulator that enables a designer to perform simulations on distant computer clusters. Performances of the developed parallel simulation algorithm are evaluated by simulation of a microelectromechanical system.     

Autonomic workflow execution in the grid

Mobile agents are being leveraged in both workflow management and grid computing contexts. The convergence of these two research streams supports execution in the grid where tasks are allowed to vary in their level of interdependence. The result is an expansion of grid applications beyond those which consist of homogeneous computations decomposed and performed in parallel to those which support the parallel execution of sequences of interdependent tasks that constitute a workflow. However, grid computation of critical workflows requires that the grid platform exhibits the autonomic characteristic of self-healing in order to ensure workflow execution. To address this issue, in this work, we first develop a model for dynamic fault tolerance technique selection, which can be embedded generically in a mobile agent workflow management system. We then augment an existing architecture for flexible fault tolerance in the grid with our model, thus allowing the system to optimally configure its fault tolerance mechanisms through awareness of the computational environment. The result is a foundation for autonomic workflow management in the grid.     

基于高速局域网的计算机制全息并行计算算法及实现

针对计算机制全息用于三维显示实用化的一个关键问题——计算速度问题，在深入分析了计算机制全息的特点基础上，提出并实现了在高速局域网中采用多台微机组成机群进行并进行计算的算法。文中还对算法的并行性、容错性及效率进行了讨论。     

基于并行计算模型的并行微粒群算法的性能分析

将微粒群算法与并行计算模型相结合,基于三种不同的并行计算模型(带中央控制器的并行计算模型、环形结构带缓存区的并行计算模型、BSP并行计算模型),设计出相应的并行微粒群算法,并对并行算法性能进行详细分析。大量实验表明:子种群之间的通讯周期是个重要的可变参数,当选取合适时,能提高解的质量以及算法的收敛性和最优性。     

Pervasive grid for large-scale power systems contingency analysis

Optimal control and management of power systems require extensive analyses of phenomena that can compromise their operation in order to evaluate their impact on the security and reliability levels of the electrical networks. For complex networks, this process, known as power systems contingencies analysis, requires large computational efforts, whereas computation times should be less than a few minutes for the information to be useful. Even though many architectures based on conventional parallel and distributed systems have been widely proposed in the literature, they are characterized by low extensibility, reusability, and scalability, and so, they require a sensible hardware upgrade when more computational resources are necessary. This event is not infrequent in power systems where the constant growth of the electrical network complexity and the need for larger security and reliability levels of the plant infrastructures lead to the need of more detailed contingency analysis in shorter times. To address this problem, this paper proposes a pervasive grid approach to define a user-friendly software infrastructure for data acquisition from electrical networks and for data processing in order to simulate possible contingencies in a real electrical network. The grid infrastructure adopts a brokering service, based on an economy-driven model, to satisfy the quality of service constraints specified by the user (i.e., a time deadline to simulate the contingencies). This paper also discusses the deployment of the infrastructure on a network of heterogeneous clusters and PCs to compute the contingency analysis of a realistic electrical network. The experimental results obtained demonstrate the effectiveness of the proposed solution and the potential role of grid computing in supporting intensive computations in power systems.     

FDTD并行算法实现及其数据通信优化

应用计算机局域网,采用基于消息传递（MPI）方式和区域分割技术,实现了FDTD的并行计算。以无限长线电流源在自由空间辐射为算例,对并行FDTD算法进行了验证。结果表明并行算法和串行算法计算结果一致,并有效提高了计算效率。最后通过减少通信数据量、优化数据交换方式及通信和计算重叠的方法,使并行算法的数据通信得到优化,并行计算效率明显提高。