PVM结构分析（3）——PVM监护进程与任务消息处理

本文着重探讨了ＰＶＭ监护进程在整个ＰＶＭ系统中的核心地位，并对其实现进行了由浅入深的剖析。ＰＶＭＤ的高层对来自网络上远地宿主机的报文和来自本地任务的报文进行分别处理。本文同时也介绍了ＰＶＭＤ中对用户层ｌｉｂｐｖｍ支持的任务消息处理部分。最后，指出了ＰＶＭ中的一些不足和改进方向     

自已动手架设UBB讨论区

安装ＡＣＴＩＶＥ ＰＥＲＬ ｌ由于 ＵＬＴＩＭＡＴＥ ＢＨＬＬＥＴＩＮＢＯＡＲＤ ＭＭ．Ｍ＊ｕ ＣＧＩ撰写的讨论区程序,因此如果要在ＷＩＮＤＯＷＳ环境下执行,就必须安装ＰＥＲＬ执行程序。所以首先请到这里下载免费的ＡＣＴＩＶＥＰＥＲＬ程序：ＨＴＴＰ：／／ＷＷＷ．ＡＣＴＩＶＥＳＴＡＴＥ．ＣＯＭ／ＤＯＷＮ－ＬＯＡＤ／ＡＣＴＩＶＥＰＥＲＬ／ＷＩＮＤＯＷＳ／５．６／ＡＣＴＩＶＥＰＥＲＬ－５．６．０．６２３－ＭＳＷＩＮ３２－Ｘ８６－ＭＵＬＴＩ—ＴＨＲＥＡＤ．ＭＳＩ。 下载完后,请直接执行ＡＣＴＩＶＥＰＥＲＬ－５．…     

我用KT133—大众AZ11、AZ31主板

随着ＡＭＤ使用Ｓｏｃｋｅｔ（Ｓｏｃｋｅｔ４６２）结构的集成了全速 Ｌ２ Ｃａｃｈｅ的 ＣＰＵ,也就是现在大家已经十分熟悉了的雷鸟和毒龙了,我们也先后作了几次的ＣＰＵ性能对比测试。使用ＣＰＵ自然就少不了用配合使用的主板,虽然目前原来很流行的 ＶＩＡ ＫＸ１３３芯片组不能很好地支持它们,只有ＡＭＤ自己的ＡＭＤ７５０可以使用,但是ＶＩＡ已经很快地推出了ＫＴ１３３芯片组来支持新的 ＡＭＤ ＣＰＵ。使用此芯片组的主板我们已经拿了一块,就是今天我们要向大家介绍的大众ＡＺ１１。 首先我们来看看ＡＺ１１的鸟瞰图（图 １…     

VX CAD/CAM从概念走向生产的捷径

在１５年的时间里,ＶＸ公司（以前称为Ｖａｒｉｍｅｔｒｉｘ）在ＣＡＤ／ＣＡＭ技术上达到了一个新的境界,为国际大型制造公司提供完整的解决方案。现在,ＶＸ公司带来先进的ＣＡＤ／ＣＡＭ软件──广泛地用于工程学、工业设计领域。 ＶＸ ＣＡＤ／ＣＡＭ系统独特的几何造型内核 在ＶＸ视窗内部是第二代专利技术,我们称之为ＵＰＧ２。就各方面来说,ＵＰＧ２核心及其数据库都有不错的性能。在功能、速度与稳定性的独立标准测试中,ＶＸ都优于当今的任何一款ＣＡＤ／ＣＡＭ产品（无论其价位怎样）。 ＵＰＧ２核心十分稳定的关键是依靠渐进…     

PVM任务迁移协议的研究

被作为ＭＰＩ“事实标准”的ＰＶＭ是一个优秀的异构多机分布计算环境支持软件,出于最初系统设计的考虑,它具极好的静态可伸缩性,但是它不支持负载在系统内部浮动,严重影响了系统资源利用率和并行效率的提高。因此,为ＰＶＭ增加任务迁移功能,使它成为可负载均衡的动态ＰＶＭ系统是近年来的研究热点。     

EDS UNIGRAPHICS V10．4

ＥＤＳＵＮＩＧＲＡＰＨＩＣＳＶ１０．４最新版本ＵＧＶ１０．４是目前最全面的ＣＡＤ／ＣＡＥ／ＣＡＭ系统．拥有４０多个应用软件，并能在各种平台上运行，如ＨＰ７００系列、ＶＡＸ和ＡＸＰ系列、ＳＧＩ、ＳＵＮ和ＩＢＭＲＳ／６００系列的各种机器。与老版本相比，１...     

SPMD模式PVM任务均衡分配的研究与实现

ＰＶＭ现有的任务分配算法较为简单，从而严重影响着系统的并行效率。本文在设计了ＵＮＩＸ负载采集系统ＵＬＩＣＳ之后，提出了实现ＳＰＭＤ模式ＰＶＭ任务均衡分配的算法，最后给出了实验结果和设计优点。     

异步检查点容错PVM

以工作站簇为代表的计算环境是当前分布式系统和并行计算的研究重点之一,ＰＶＭ所提供的消息传递机制支持了高效的异构网络计算。但标准ＰＶＭ缺乏对系统容错的支持,这可以通过使用检查点的回滚恢复方式予以弥补。该文对如何在用户级实现ＰＶＭ全局容错,分析其设计思想和实现技术。主要思想是使用进行消息记录的异步检查点算法,并利用ＰＶＭ守护进程和全局调度进程进行控制,所有操作对应用程序都是透明的。利用该系统还可以进一步实现ＰＶＭ的透明进程迁移和负载均衡。     

适合大众家庭使用的装机计划

Ｃ Ｐ Ｕ： Ｉｎｔｅｌ奔腾Ⅱ４００ Ｓｌｏｔ１／５１２Ｋ散装土板：升技 ＶＡ６ ＶＩＡ６９３Ａ／ＡＴＸ／Ｓｌｏ－１／支持１３３ＭＨｚ外频／支持 ＤＭＡ／６６／带声卡硬盘： ＩＢＭ ＤｅｓｋｓｔａｒＧＸＰ／ＤＰＴＡ３７２０５０ ２０．５Ｇ／ＵＤＭＡ／６６／２Ｍ／７２００ｒｐｍ／ＩＤＥ／全国联保三年内存： ＨＹ现代／１２８Ｍｂ ７ｎｓ／１６８ＰＩＮ ＳＤＡＭ／ＰＣ １００显示卡：ＡＴＩ ＸＰＥＲＴ ２０００Ｒａｇｅ　１２８ＧＬ／３２ＭＢＳＤＲＡＭ／ＡＧＰ２Ｘ声卡：主板载有 ＡＣ’９７ Ａｕｄｉｏｎ Ｃｏｄｅｃ支持１…     

装台让自己可以炫耀一番的机子

Ｃ　Ｐ Ｕ： Ｉｎｔｅｌ赛扬 ３６６ Ｓｏｃｋｅｔ３７０／１２８Ｋ／散主板：升技 ＶＡ６ ＶＩＡ ６９３Ａ／ＡＴＸ／Ｓｌｏ－１／支持 １３３ＭＨｚ外频／支持 ＤＭＡ／６６／带声卡硬盘： ＩＢＭ ＤｅｓｋＳｔａｒＧＸＰ／ＤＰＴＡ３７２０５０ ２０．５Ｇ／ ＵＤＭＡ／６６／２Ｍ／７２００ｒｐｍ／ＩＤＥ／全国联保三年内存：ＨＹ现代／１２８Ｍｂ ７ｎｓ／１６８ＰＩＮ　ＳＤＲＡＭ／ＰＣ　１００ 显示卡：帝盟 Ｖｉｐｅｒ Ｖ７７０ Ｕｌｔｒａ Ｒｉｖａ ＴＮＴ２ Ｕｌｔｒａ／３２Ｍ ＳＤＲＡＭ／ＡＧＰ ２ｘ／４ｘ声卡：主板载有 Ａ…     

基于网络的分布并行虚拟计算机的实现

构造了一个基于网络的分布并行虚拟计算机DPVM,它由虚拟机层、通信层和基本类层组成,包括服务器,工作机和客户机3种不同类型的机器,在简要介绍DPVM的总体结构之后,从服务器,任务通信,程序输出和消息传递并行语义的面向对象实现几方面对系统实现的关键技术进行讨论。     

Object Mobility for Performance Improvements of Parallel Java Applications

This paper describes a framework that employs Java and mobile objects to provide an efficient, parallel-computing environment in a heterogeneous, scalable computing cluster. The framework consists of migratable active objects performing computational tasks and static objects for managing them. The proposed framework simplifies the development of parallel applications by providing means for transparent object distribution among the nodes. It also improves the overall run-time performance by efficient, cluster-wide resource utilization. The framework uses adaptive resource management algorithms which are particularly attractive for executing parallel programs with unpredictable resource requirements and in clusters with heterogeneous machines. Our experiments show a significant improvement in the execution time of parallel applications when using the proposed schemes.     

曙光超级服务器上DMPI和DPVM的设计

介绍了运行于曙光超级服务器上的DMPI和DPVM的设计与实现。DMPI支持RMA和动态进程管理的功能。DMPI和DPVM在对ADI层进行灵活有效的扩展(DADI-E)的基础上提供了一个高性能的通信层。该文对DADI-E的特点进行了阐述,包括RMA机制、动态进程管理和对PVM的支持。而且,DMPI还提供了一个特殊的数据通路———流水发送,它可以改善DMPI的带宽,使之接近于底层通信协议的带宽峰值。最后还给出了DMPI和DPVM在Myinet卡上的性能指标。     

An experimental evaluation of a peer-model monitoring system for the support of a parallel processing environment

The process of monitoring the machines or computing nodes in a network, and of monitoring the communication traffic between them, is very important to efficiently launch and execute parallel coarse-grained applications or even classical (serial-type) applications, taking advantage of machines in the network that are not heavily used. An experimental software system, named MONSYS, that is capable of monitoring the machines in a network, is presented. MONSYS can be used in the support of an Application Manager system capable of distributing parallel tasks (or classical programs) over the machines in a local area network with the objective of achieving load balancing. It can also be used as a tool in the administration of networks. MONSYS exhibits a highly decentralized and fault tolerant architecture based on the Peer-Model, which, together with its information diffusion algorithms, constitute its prime novelty. A set of experiments that investigate the performance and scalability of a prototype of MONSYS is presented and discussed. The experiments reported show that MONSYS offers a reasonably accurate picture of the internal state of the machines monitored, without being a burden to the network communication channels or to the machines themselves. In fact, the quantitative results obtained indicate that MONSYS can be several times more performant than an equivalent system using a multicast communication scheme for the exchange of machine state information.     

Distributed versus parallel computing

The elegant but simple von Neumann single processing design of computers has been challenged by new applications of databases, computer vision and speech where a multi-processing system seems more suited to such tasks. We look at three ways in which parallel machines may be used: for general purpose computing, for algorithms which are not well suited to von Neumann machines and for exploring forms of computation which cannot reasonably be dealt with on von Neumann machines.     

基于网络的分布并行虚拟计算机DPVM的总体设计

基于网络的分布式并行计算技术提供高性能、低成本的计算资源。该文介绍所构造的一个基于网络的分布并行虚拟计算机DPVM。它由虚拟机层、通信层和基本类层组成,以浏览器方式提供用户界面;允许计算机加入系统,提供空闲计算资源,或登录到系统上,获得计算资源;采用Java作为程序设计语言,支持平台无关的分布并行计算。     

Sensitivity bounds for machine scheduling with uncertain communication delays

The paper deals with the scheduling of tasks on a set of parallel machines subject to precedence constraints and communication delays, that are not fully known before the execution. This is usual in parallel computing (message passing) and in production workshops (part transportation). The goal is to give a priori informations on the behavior of different schedules computed before the execution. This sensitivity analysis helps to choose among different available algorithms. It also provides tools to decide whether it is justified to change the schedule structure at execution time (run-time scheduling).By hypothesis, the communication delays are within a given interval. Worst case sensitivity bounds are proposed that give the maximum loss of performance. These bounds hold for algorithms which are optimal in the deterministic case, but can be easily adapted to arbitrary algorithms. Some applications to specific problems are presented, namely two identical machines, and unbounded number of machines, for tree-like precedence constraints.     

一种面向虚拟化数字中心资源按需重配置方法

面向Web应用,提出一种动态资源按需配置方法,能够根据不断变化的资源需求以在线方式高效地重配置集群,实时地确定集群当前节点运行数量及其上部署的虚拟机类型.该方法基于布尔二次指数平滑法预测用户请求,有效避免了配置结果落后于资源请求;基于遗传算法并行化搜索配置空间,快速发现合理配置.实验结果表明,该方法能够根据需求变化高效地在线调整系统资源配置,并可有效提高集群资源利用率,显著降低了系统能耗.     

Parallel approaches to machine learning—A comprehensive survey

Literature has always witnessed efforts that make use of parallel algorithms / parallel architecture to improve performance; machine learning space is no exception. In fact, a considerable effort has gone into this area in the past fifteen years. Our report attempts to bring together and consolidate such attempts. It tracks the development in this area since the inception of the idea in 1995, identifies different phases during the time period 1995–2011 and marks important achievements. When it comes to performance enhancement, GPU platforms have carved a special niche for themselves. The strength of these platforms comes from the capability to speed up computations exponentially by way of parallel architecture / programming methods. While it is evident that computationally complex processes like image processing, gaming etc. stand to gain much from parallel architectures; studies suggest that general purpose tasks such as machine learning, graph traversal, and finite state machines are also identified as the parallel applications of the future. Map reduce is another important technique that has evolved during this period and as the literature has it, it has been proved to be an important aid in delivering performance of machine learning algorithms on GPUs. The report summarily presents the path of developments.     

Simulation-Based Performance Prediction for Large Parallel Machines

We present a performance prediction environment for large scale computers such as the Blue Gene machine. It consists of a parallel simulator, BigSim, for predicting performance of machines with a very large number of processors, and BigNetSim, which incorporates a pluggable module of a detailed contention-based network model. The simulators provide the ability to make performance predictions for very large machines such as Blue Gene/L. We illustrate the utility of our simulators using validation and prediction studies of several applications using smaller numbers of processors for simulations.