Asynchronous Transfer Mode and other Network Technologies for Wide-Area and High-Performance Cluster Computing

We review fast networking technologies for both wide-area and high performance cluster computer systems. We describe our experiences in constructing asynchronous transfer mode (ATM)-based local- and wide-area clusters and the tools and technologies this experience led us to develop. We discuss our experiences using Internet Protocol on such systems as well as native ATM protocols and the problems facing wide-area integration of cluster systems. We are presently constructing Beowulf-class computer clusters using a mix of Fast Ethernet and Gigabit Ethernet technology and we anticipate how such systems will integrate into a new local-area Gigabit Ethernet network and what technologies will be used for connecting shared HPC resources across wide-areas. High latencies on wide-area cluster systems led us to develop a metacomputing problem-solving environment known as distributed information systems control world (DISCWorld). We summarize our main developments in this project as well as the key features and research directions for software to exploit computational services running on fast networked cluster systems.     

计算网格中多编程环境的研究与实现

计算网格(也称元计算系统)聚集地理上分散的资源进行大型的分布式高性能计算。PVM和MPI是广泛使用的并行编程环境,它们需要作为并行计算的基本构建而集成到元计算系统中去。论文针对元计算资源的动态性、分布性、性能多变性和结点异构性等特点,实现了一个自适应的、一体化的多编程环境。论文论述了该多编程环境的体系结构,并利用代理技术实现远程编译、发现资源、屏蔽异构和优化调度。     

基于MDI的异构数据格式转换和远程文件传输机制研究

元计算系统的节点间具有多方面的异构性，当数据在这些点间传输时，会发生理解或判断上的歧义性，为了实现数据和文件在异构节点间传输的一致性，分析了校园网络内的几种典型计算机系统的数据类型及其存储格式，研究了不同数据格式之间的转换算法和远程文件传输机制，格式转换算法使用元数据信息作为数据信包转换和解压缩的依据。     

基于高速网络的广域高性能并行与分布式计算

本文试图说明以下观点：越来越多的高性能应用要求利用地理上分布的、各式各样的计算和数据资源。这些应用希望能够通过高速网络将地理上分布、异构的各种高性能计算机、数据服务器、大型检索存储系统和可视化、虚拟现实系统等连接并集成起来，形成一个网络虚拟计算机（称为元计算机），来实现应用计算问题。这种元计算实质上就是基于高速网络的广域高性能并行与分布式计算。     

JiniSolve:一种基于Jini的元计算架构的基本体系结构及其初步应用

随着不同种类体系结构高性能计算机系统的不断出现和网络技术的迅猛发展,人们越来越希望能够打破平台和地域的限制,共享分布在广域网上的各种高性能计算机的计算能力。但广域网络分布式计算平台的异构性和资源组织协调的复杂性一直困扰着这方面工作的发展。作者从Java字节码的平台无关性和Jini架构的网络资源运行时动态自组织能力中得到启发,提出了JiniSolve网络异构平台分布式计算架构(一种元计算架构),以克服这两个障碍。JiniSolve架构能够支持所有三种网络分布式计算的模式———远程计算、MobileAgent计算和代码下载计算,其中对MobileAgent计算的支持是以往类似元计算架构所不能做到的。同时,JiniSolve架构又具有运行时动态异步自组织能力,从而使整个系统具有更高的可适应性,可用性,可靠性和易用性。文章将介绍JiniSolve架构的基本体系结构,分析其优点并给出一个实验性的应用举例。     

基于构件技术的元计算环境框架的任务分配与调度

本文从软件复用的角度对元计算环境进行了研究，提出了一个基于构件技术的元计算环境框架MefBC，并重点给出了基于此框架任务的分配与调度．在一般的元计算中，任务结点在其所有前驱结点执行结束后才能进行分配调度．本文则采用了动态调度方案，任务结点是根据其前驱结点的执行情况来分配调度到下一个计算结点的，并且任务结点的分配调度在其一前驱结点执行完毕时就开始，以实现计算跟通信的最大重叠，提高系统效率．     

Collaborative Metacomputing with IceT

The IceT metacomputing system is based on the principles of collaborative resource sharing, lightweight software frameworks, and efficient execution in heterogeneous environments. In this project, the traditional model for distributed computing environments has been extended to facilitate the controlled and secure exchange of hardware and software resources, and to support programming paradigms that achieve a balance between portability and efficiency. The current IceT system is Java-based, and presents a framework within which non-persistent collections of resources, users, and processes may interact to accomplish distributed problem solving. This paper describes the architecture and design principles of IceT, its operational model and support for the execution of distributed applications, and orthogonal subsystems that were developed to support the concept of cooperative metacomputing.     

Solving Large MINLPs on Computational Grids

We consider the solution of Mixed Integer Nonlinear Programming (MINLP) problems by a parallel implementation of nonlinear branch-and-bound on a computational grid or meta-computer. Computational experience on a set of large MINLPs is reported which indicates that this approach is efficient for the solution of these problems.     

Using multicast communications to distribute code and data in wide area networks

This paper describes the implementation and use of a prototype multicast late-binding RPC (LbRPC) system. The system provides a mechanism for exporting arbitrary code and data across a wide area internetwork to multiple hosts for evaluation. This mechanism provides a means to ensure that the total execution time for the application is as small as possible without requiring continual monitoring and estimation of the state of the hosts and the intervening network. The impact of sending the same code and data to a group of remote hosts is minimized by using multicast Internet protocol (IP) communications for the outward leg of the transaction. This prototype implementation demonstrates the feasibility of using multicasting for process placement and some performance measurements are included to show how it stands up to usage on the Internet. The paper concludes with a discussion of outstanding issues raised by this research.     

元计算系统的一个可扩展层次型资源管理模型

资源管理是元计算系统中的关键问题，为了支持大规模元计算应用，必须有效地解决资源联合分配问题。提出了一种可扩展层次型资源管理模型，在充分利用已有的批处理调度系统的基础上，通过核心资源管理器与局域资源调度系统的交互，为异构资源提供统一的接口，为解决资源的联合分配问题，提出了基于站点的层次型管理机制，可将应用程序的资源请求分布在相对紧密的资源集合上，不仅减小了元计算应用程序的整体通信开销，而且计算提供了资源预约服务。