A Compute Unified System Architecture for Graphics Clusters Incorporating Data Locality

We present a development environment for distributed GPU computing targeted for multi-GPU systems, as well as graphics clusters. Our system is based on CUDA and logically extends its parallel programming model for graphics processors to higher levels of parallelism, namely, the PCI bus and network interconnects. While the extended API mimics the full function set of current graphics hardware—including the concept of global memory—on all distribution layers, the underlying communication mechanisms are handled transparently for the application developer. To allow for high scalability, in particular for network-interconnected environments, we introduce an automatic GPU-accelerated scheduling mechanism that is aware of data locality. This way, the overall amount of transmitted data can be heavily reduced, which leads to better GPU utilization and faster execution. We evaluate the performance and scalability of our system for bus and especially network-level parallelism on typical multi-GPU systems and graphics clusters.     

Incorporating negative values into the Analytic Hierarchy Process

This paper describes why and how ratio scaled multiple criteria analysis techniques, such as the Analytic Hierarchy Process (AHP), should allow for a subtraction mechanism whereby negative preference (aversion) is combined with positive preference. The main contribution of the paper is a demonstration of how the current imposition of a strictly positive additive value structure can lead to incorrect preference ratios and even incorrect ranking of alternatives. The proposed preference elicitation and computation method solves these problems in a simple and intuitive manner.

Scope and purpose

We focus in this paper on common cases where the AHP is used to evaluate alternatives in light of multiple criteria and where at least some of the considerations involve negative rather than positive effects from the point of view of the decision maker. The main objective is to correct how such negative effects are evaluated within the AHP.     

Integrating visual goal models into the Rational Unified Process

The Rational Unified Process is a comprehensive process model that is tailorable, provides templates for the software engineering products, and integrates the use of the Unified Modeling Language (UML); it is rapidly becoming a de facto standard for developing software. The process supports the definition of requirements at multiple levels. Currently, the early requirements, or goals, are captured in a textual document called the Vision Document, as the UML does not include a goal modeling diagram. The goals are subsequently refined into software requirements, captured in UML Use Case Diagrams. Given the well documented advantages of visual modeling techniques in requirements engineering, including the efficient communication and understanding of complex information among numerous diverse stakeholders, the need for an enhanced version of the Vision Document template which supports the visual modeling of goals is identified. Here, an Enhanced Vision Document is proposed which integrates two existing visual goal models: AND/OR Graph for functional goals and Softgoal Interdependency Graph for non-functional goals. A specific approach to establishing traceability relationships from the goals to the Use Cases is presented. Tool support has been developed for the Enhanced Vision Document template; the approach is illustrated using an example system called the Quality Assurance Review Assistant Tool.     

A network user interface: Incorporating human factors guidelines into the ISO standard for open systems interconnection

The introduction of standards will hopefully ensure that users can access particular computer resources through a communications network for their own purposes without major problems. The International Standards Organisation (ISO) has developed a seven-layer reference model which is to be used for the purpose of incorporating standards relating to the interconnection of open systems (OSI). It is important that the human factors requirements are considered in relation to this model if the aim of generality of use is to be achieved. This paper considers some of the major human factors requirements and describes an approach to translating them into design standards which can be implemented. The approach starts from a consideration of user activity and develops into a language interface which could reside in layers of the reference model.     

基于UML的C4ISR体系结构设计过程

本文以C4ISR体系结构框架2.0为体系结构的描述规范,在分析C4ISR体系结构和我军军事信息系统体系结构的基础上,总结了体系结构设计的主要技术与方法,结合C4ISR体系结构的通用开发过程,提出了基于统一建模语言(UML)的C4ISR体系结构设计过程.并较深入地探讨了在C4ISR体系结构设计过程中用统一建模语言(UML)中的图表进行开发的详细步骤,以及体系结构产品的开发流程.     

基于UML的C^4ISR体系结构设计过程

本文以C^4ISR体系结构框架2．0为体系结构的描述规范，在分析C^4ISR体系结构和我军军事信息系统体系结构的基础上．总结了体系结构设计的主要技术与方法，结合C^4ISR体系结构的通用开发过程，提出了基于统一建模语言（UML）的C^4ISR体系结构设计过程。并较深入地探讨了在C^4ISR体系结构设计过程中用统一建模语言（UML）中的图表进行开发的详细步骤，以及体系结构产品的开发流程。     

一元化安全结构初探

该文分析了企业网络的安全需求及其采用的安全策略的不足,阐述了什么是一元化安全结构,一元化安全结构的好处,及目前一元化安全结构的应用。     

基于进程组构件模型的软件体系结构设计方法

软件工程的有效性在于进行体系化或系统化的设计,软件体系结构的好坏对系统的成败起着举足轻重的作用。本文针对并发系统,结合面向对象、构件以及软件复用的思想,提出了一种基于进程组构件模型的软件体系结构设计方法。最后介绍了采用该方法设计的iCALL系统。     

NVIDIA Tesla: A Unified Graphics and Computing Architecture

To enable flexible, programmable graphics and high-performance computing, NVIDIA has developed the Tesla scalable unified graphics and parallel computing architecture. Its scalable parallel array of processors is massively multithreaded and programmable in C or via graphics APIs.     

A comparison of the ANSI and ISO Pascal Standards

We compare the American and international standards for Pascal and make suggestions to help implementors and programmers achieve the greatest possible compatibility with both standards.     

Designing and Rightsizing the Information System Architecture

Information system design and sizing constitute a complex, top-down process that tailors the technology architecture to application requirements. Practitioners usually tackle this top-down process by focusing on individual design aspects, such as data or specific applications, and by relying on their previous experiences to compare alternative architectural solutions. Acquisition costs are usually accounted for, but related operating and maintenance costs are often neglected or underestimated. The complexity of optimizing individual design problems leads researchers to avoid a global optimization perspective and, thus, the IS architecture is usually a result of the juxtaposition of multiple local optima.This paper takes an overall perspective on the cost minimization problem of information system design to achieve a better trade-off between cost and performance over the whole expected life of the technology architecture. A comprehensive design methodology is discussed as an integrating framework that accounts for all categories of costs, including design, implementation, maintenance, and operation, to achieve a globally cost-minimizing solution.     

Data Architecture Standards for the Effective Enterprise

The modern enterprise has become increasingly dependent on data and its value-added forms of information and knowledge to remain competitive in the face of global competition and constant change. Data architecture provides the framework necessary to use and share data more effectively, and to improve the flow of data between systems within the enterprise and between the enterprise's systems and those of its customers, suppliers, and business partners. Data architecture standards constitute the foundation of an effective data architecture.     

统一软件开发过程述评

1 引言统一软件开发过程(Unified Software DevelopmentProcess,USDP)是由统一建模语言(Unified ModelingLanguang,UML)的缔造者们提出来的.并为对象管理组织(Object Management Group,OMG)所推荐。统一软件开发过程是缔造者们权衡了三十年的软件开发实践而形成的产物。它把当前处于分离的方法学集合于一起,称之为“统一方法”。它不仅统一了缔造者们的工作,而且使得开发方法得以统一,方法学家的工作得以统一。它也包含了很多其他人员和公司对UML作出的巨大贡献,例如它来自于数百个用户多年的现场经验以及Rational公司多年的工作成果。 USDP对于如何运用UML的概念进行软件开发提供了详细指导,它指导开发队伍安排其开发活动的次序,为各开发     

软件过程与统一软件开发过程研究

论文介绍了软件过程与软件建模技术的发展历史,讨论了软件过程、建模语言和统一软件开发过程的基本概念和原理,并重点讨论了软件构架、用例驱动、迭代与增量等技术的内涵。尽快掌握并适应这种新方法,开发与之配套的软件环境与工具是当务之急,这对于软件产业和软件工程学科有重要的意义和作用。     

PESOI： Process Embedded Service-Oriented Architecture

1 Introduction Service-Oriented Architecture (SOA) has received significant attention recently as major computer and software companies such as HP, IBM, Intel, Microsoft, and SAP, have all embraced SOA, as well as government agencies such as DoD (US depar…     

企业MIS建设和ISO标准化建设的融合

从企业管理信息系统建设和企业标准化建设的过程和特点,分析论述了企业两者建设之间的关系并说明只有两者融合建设才能经企业带来巨大的成功。     

Know means no: Incorporating knowledge into discrete-event controlsystems

Modal logic is introduced into the modeling of discrete-event systems. Analysis within this framework includes formal reasoning about what supervisors know or do not know about a given system. This model can be used to develop control strategies that solve decentralized discrete-event control problems. When a problem cannot be solved using fully decentralized supervisors, reasoning about knowledge may provide guidelines for incorporating communication and pooled information into the model     

Incorporating AI into military decision making: an experiment

The Integrated Course of Action Critiquing and Elaboration System integrates several available technologies based on AI techniques. These techniques range from machine-understandable course-of-action representations entered via sketching and constrained natural language to interleaved adversarial planning and scheduling.     

Incorporating fuzzy membership functions into the perceptron algorithm

The perceptron algorithm, one of the class of gradient descent techniques, has been widely used in pattern recognition to determine linear decision boundaries. While this algorithm is guaranteed to converge to a separating hyperplane if the data are linearly separable, it exhibits erratic behavior if the data are not linearly separable. Fuzzy set theory is introduced into the perceptron algorithm to produce a ``fuzzy algorithm' which ameliorates the convergence problem in the nonseparable case. It is shown that the fuzzy perceptron, like its crisp counterpart, converges in the separable case. A method of generating membership functions is developed, and experimental results comparing the crisp to the fuzzy perceptron are presented.     

多出口校园网体系结构分析设计

近年来,随着学校信息化建设工作的深入开展,对校园网的要求也越来越高,传统的单一外网接入模式已经很难满足日趋复杂的应用需求。目前,大多数学校都采取了多个ISP同时接入校园网的方式。如何设计好多出口校园网体系结构,使其提供高质量的服务,成了摆在面前的现实问题。本文从不同的角度出发,对多出口校园网体系结构进行分析,并针对其存在的一些常见的问题给出了对应的解决办法。