Technique Development and Application——Construction of a Beowulf Cluster for Parallel Computing

The large-scale computations are often performed in science and engineering areas such as numerical weather forecasting, astrophysics, energy resources exploration, nuclear weapon design, and plasma fusion research etc. Many applications in these areas need super computing power. The traditional mode of sequential processing cannot meet the demands of those computations, thus, parallel processing（PP） is the main way of high performance computing （HPC） now.

Heuristic file sorted assignment algorithm of parallel I/O on cluster computing system

A new file assignment strategy of parallel I/O, which is named heuristic file sorted assignment algorithm was proposed on cluster computing system. Based on the load balancing, it assigns the files to the same disk according to the similar service time. Firstly, the files were sorted and stored at the set I in descending order in terms of their service time, then one disk of cluster node was selected randomly when the files were to be assigned, and at last the continuous files were taken orderly from the set I to the disk until the disk reached its load maximum. The experimental results show that the new strategy improves the performance by 20.2% when the load of the system is light and by 31.6% when the load is heavy. And the higher the data access rate, the more evident the improvement of the performance obtained by the heuristic file sorted assignment algorithm.

Parallel computing for finite element structural analysis using conjugategradient method based on domain decomposition

Parallel finite element method using domain decomposition technique is adapted to a distributed parallel environment of workstation cluster. The algorithm is presented for parallelization of the preconditioned conjugate gradient method based on domain decomposition. Using the developed code, a dam structural analysis problem is solved on workstation cluster and results are given. The parallel performance is analyzed.

PARALLEL ANALYSIS OF COMBINED FINITE/DISCRETE ELEMENT SYSTEMS ON PC CLUSTER

A computational strategy is presented for the nonlinear dynamic analysis of large- scale combined finite/discrete element systems on a PC cluster.In this strategy,a dual-level domain decomposition scheme is adopted to implement the dynamic domain decomposition.The domain decomposition approach perfectly matches the requirement of reducing the memory size per processor of the calculation.To treat the contact between boundary elements in neighbouring subdomains,the elements in a subdomain are classified into internal,interfacial and external elements.In this way,all the contact detect algorithms developed for a sequential computation could be adopted directly in the parallel computation.Numerical examples show that this implementation is suitable for simulating large-scale problems.Two typical numerical examples are given to demonstrate the parallel efficiency and scalability on a PC cluster.

Parallel Computation of Shallow-water Model on Workstations Cluster

ＰａｒａｌｌｅｌＣｏｍｐｕｔａｔｉｏｎｏｆＳｈａｌｌｏｗ－ｗａｔｅｒＭｏｄｅｌｏｎＷｏｒｋｓｔａｔｉｏｎｓＣｌｕｓｔｅｒＳｏｎｇＪｕｎｑｉａｎｇ；ＳｕｎＡｎ－ｃｌａｎｇ，；ＬｉＸｉａｏｍｅｉ（ｅｐａｒｔｍｅｎｔＯｆＣＯ，ｍｐ＇ｕｔｅｒＳｃｉｅｎｃｅ，...

LOW-COST HIGH PERFORMANCE CLUSTER OF WORK-STATIONS BASED ON DYNAMIC LOAD BALANCING FOR PARALLEL DEPTH-FIRST SEARCH(DFS)

This paper presented an idea to replace the traditionally expensive parallel machines by heterogeneous cluster of workstations. To emphasise the usability of cluster of workstations platform for parallel and distributed computing, also the paper presented the status report on the effort and experiences for the implementation of a dynamic load balancing for parallel tree computation depth first search(DFS) on the cluster of a workstations project. It compared the speedup performance obtained from our platform with that obtained from the traditional one. The speedup results show that cluster of workstations can be a serious alternative to the expensive parallel machines.

Implementation Study of Dynamic Load Balancing Algorithm of Parallel Tree Computation on Clusters of Heterogeneous Workstation

The rapid growth of interconnected high performance workstations has produced a new computing paradigm called clustered of workstations computing. In these systems load balance problem is a serious impediment to achieve good performance. The main concern of this paper is the implementation of dynamic load balancing algorithm, asynchronous Round Robin (ARR), for balancing workload of parallel tree computation depth-first-search algorithm on Cluster of Heterogeneous Workstations (COW) Many algorithms in artificial intelligence and other areas of computer science are based on depth first search in implicitty defined trees. For these algorithms a load-balancing scheme is required, which is able to evenly distribute parts of an irregularly shaped tree over the workstations with minimal interprocessor communication and without prior knowledge of the tree’s shape. For the (ARR) algorithm only minimal interprocessor communication is needed when necessary and it runs under the MPI (Message passing interface) that allows parallel execution on heterogeneous SUN cluster of workstation platform. The program code is written in C language and executed under UNIX operating system (Solaris version).

Targeting BSP Library for SMP Cluster

Using commodity SMPs (shared memory processors) to build cluster-based supercomputer has become a mainstream trend.Yet programming this kind of supercomputer system requires an environment support both message passing and shared memory programming. This paper describes our preliminary work in an effort to target BSP library for cluster of SMPs. In order to exploit the maximum performance potential that a cluster of SMPs brings, we adopt thread technique to reduce system overhead and to exploit the capacity of SMPs. A fore-layer synchronization mechanism is proposed to support barrier synchronization within an SMP node, a group of SMP nodes and the whole cluster respectively. A comparison is made between our BSP library and the currently available BSP libraries such as PUB.

基于Linux Cluster的并行CMAQ研究

CMAQ（Community Multiscale Air Quality）涉及海量空间数据、复杂的处理模型和苛刻的时间需求，但高密集的计算操作使得串行CMaQ面临计算瓶颈问题，昂贵的巨型高性能专用机对于普通研究者望尘莫及，因此基于Linux Cluster的并行CMAQ研究是解决该问题的重要途径。本文以开源CMAQ为研究对象，探讨基于Linux Cluster的并行CMAQ的计算模式、体系结构、并行模式、软件框架等，并构建了相应的原型系统。实验表明相对于传统的串行架构，所提出的并行架构在计算效率上有了显著提高。

PC Cluster系统无法运行JSS作业的原因分析及解决方法

本文在简要介绍PC Cluster系统组成的基础上,着重分析了系统不能运行JSS作业的原因——inode被占满了而引起的,并介绍了通过删除mailman软件包排除故障的方法。

PC Cluster存储方案的设想与实施

本文简要介绍了PC Cluster系统组成,并在对几种存储系统进行分析的基础上,对SAN存储技术在PC Clus-ter系统中的实施和应用进行了详细的说明。

Efficient parallel implementation of a density peaks clustering algorithm on graphics processing unit

The density peak （DP） algorithm has been widely used in scientific research due to its novel and effective peak density-based clustering approach. However, the DP algorithm uses each pair of data points several times when determining cluster centers, yielding high computational complexity. In this paper, we focus on accelerating the time-consuming density peaks algorithm with a graphics processing unit （GPU）. We analyze the principle of the algorithm to locate its computational bottlenecks, and evaluate its potential for parallelism. In light of our analysis, we propose an efficient parallel DP algorithm targeting on a GPU architecture and implement this parallel method with compute unified device architecture （CUDA）, called the ‘CUDA-DP platform＇. Specifically, we use shared memory to improve data locality, which reduces the amount of global memory access. To exploit the coalescing accessing mechanism of CPU, we convert the data structure of the CUDA-DP program from array of structures to structure of arrays. In addition, we introduce a binary search-and-sampling method to avoid sorting a large array. The results of the experiment show that CUDA-DP can achieve a 45-fold acceleration when compared to the central processing unit based density peaks implementation.

Preliminary study of a cluster-based open-source parallel GIS based on the GRASS GIS

In response to the problem of how to give geographic information system(GIS)high-performance capabilities for certain specific GIS applications,a new GIS research direction,parallel GIS processing,has emerged.However,traditional research has focused mostly on implementing typical GIS parallel algorithms,with little discussion of how to parallelize an entire GIS package on clusters based on theory.Therefore,the authors have chosen the geographic resources analysis support system(GRASS)GIS as the object of their research and have put forward the concept of a cluster-based open-source parallel GIS(cluster-based OP-GIS)as a tool to support Digital Earth construction.The related theory includes not only the parallel computing mode,architecture,and software framework of such a system,but also various parallelization patterns.From experiments on the prototype system,it can be concluded that the parallel system has better efficiency and performance than the conventional system on certain selected modules.

Realistic Efficiency Evaluations for Parallel Computations under Workstation Cluster

In recent years, high performance scientific computing under workstation cluster connected by local area network is becoming a hot point. Owing to both the longer latency and the higher overhead for protocol processing compared with the powerful single workstation capacity, it is becoming severe important to keep balance not only for numerical load but also for communication load, and to overlap communications with computations while parallel computing. Hence,our efficiency evaluation rules must discover these capacities of a given parallel algorithm in order to optimize the existed algorithm to attain its highest parallel efficiency. The traditional efficiency evaluation rules can not succeed in this work any more. Fortunately, thanks to Culler's detail discuss in LogP model about interconnection networks for MPP systems, we present a system of efficiency evaluation rules for parallel computations under workstation cluster with PVM3.0 parallel software framework in this paper. These rules can satisfy above acquirements successfully. At last, two typical synchronous,and asynchronous applications are designed to verify the validity of these rules under 4 SGIs workstations cluster connected by Ethernet.

Generative computer simulation for parallel layout of multi-story row house cluster

This study has established the functions between the environmental conditions and the inhabitants’ preferences for multi-story row house cluster with parallel layout, based on the data from the questionnaires’ in Beijing. A program has been written with multi-agent system and generative computer simulation approaches. The emergence candidate layout plans can be referenced and chosen by architects.

CLUSTER OF WORKSTATIONS BASED ON DYNAMIC LOAD BALANCING FOR PARALLEL TREE COMPUTATION DEPTH-FIRST-SEARCH

The real problem in cluster of workstations is the changes in workstation power or number of workstations or dynmaic changes in the run time behavior of the application hamper the efficient use of resources. Dynamic load balancing is a technique for the parallel implementation of problems, which generate unpredictable workloads by migration work units from heavily loaded processor to lightly loaded processors at run time. This paper proposed an efficient load balancing method in which parallel tree computations depth first search (DFS) generates unpredictable, highly imbalance workloads and moves through different phases detectable at run time, where dynamic load balancing strategy is applicable in each phase running under the MPI(message passing interface) and Unix operating system on cluster of workstations parallel platform computing.

一种基于敏捷集群计算系统的并行GMRES方法

随着通信系统和人工智能的飞速发展,以智慧城市、智慧工厂和智能制造等为代表的多种新型应用场景不断涌现,使得通信、感知和计算等系统的一体化成为技术发展的新趋势。人工智能新型应用场景对大规模高效敏捷计算提出了新的要求,基于敏捷集群计算系统,提出了一种并行广义最小残差(Generalized Minimal Residual, GMRES)方法,主要通过并行矩阵向量乘法和并行高瘦矩阵QR(Tall and Skinny QR,TSQR)分解实现Krylov子空间的高效并行构造,充分利用集群计算系统的计算和通信性能,实现大规模线性方程组Ax=b的快速求解,其中A为一个n×n的矩阵,在工程实践中,n可达数十万甚至百万规模。通过求解二维泊松方程的有限元离散得到的刚度方程,验证了算法的有效性。

异构并行计算下高维混合型数据聚类算法研究

高维数据维度增加,数据空间的体积呈指数增长,容易陷入“维数灾难”,导致聚类算法执行效率低,为此,提出异构并行计算下高维混合型数据聚类算法。构建高维混合型数据相异度矩阵,提取高维混合型数据的统计序列特征值,利用时间窗口进行特征优化。采用K⁃Prototypes聚类算法提取高维混合型数据的统计序列特征,评估数据与类中心的相异性,计算数据与类中心的欧氏距离,实现高维混合型数据聚类。采用异构并行计算技术进行高维混合型数据K⁃Prototypes聚类的并行化处理,合理分配CPU与GPU工作,达到CPU与GPU的工作负载平衡,提高K⁃Prototypes的聚类效率。实验结果表明,此算法对于高维混合型数据的聚类效果好、运行时间短、性能稳定。

Cluster计算能力模型总结与设想

Cluster计算,中文名称为集群计算。它可以被认为是这样一种技术:它把许多系统连接到一起,使多台服务器像一台机器一样进行计算工作。采用cluster计算,通常是为了提高系统稳定性的同时,提高系统对大数据的实时处理能力。在cluster计算中,服务器群在网络中表现为单一的系统,并以这种模式加以管理。对于并行计算和集群计算,现已提出的各种计算模型,大多针对的是并行计算与它的共性,而非特性。而如果需要针对这些特性,就需要用一种更加准确的模型加以描述。将介绍基于集群的一些计算模型,并提出新的cluster计算模型的一个可能的方案。