Resource Encapsulation in Manufacturing Grid Environment
Environments for Robotics, Automation and Manufacturing) at the Software Automation and
Intelligent Coordinating Entities Based Control Software Design*S. RamaswamySoftware Automation and Intelligence LaboratoryDepartment of Computer Science Tennessee Technological UniversityCookeville TN 38505Phone: 931-372-3448Email: srini@ / srini@*Several students in SAIL have contributed to, and benefited from, research activities at the SAIL laboratory at Tennessee Technological University. These include: (i) Thad Scalf, Joe Cherry and Karthigan Srinivasan: Naval Radar Simulation Project.(ii) Andrew Trent and William Holcomb: Self Organizing Software Entities Project. (iii) Matt Irwin, Noah Rosser, Derrick Muncy and : Home Automation Project. (iv) VijayAnand: Power Plant Distribution Visualization Project.Abstract: Software systems designed to be distributed, interactive and intelligent (in a domain-specific sense), with ubiquitous human interfaces and the ability to exhibit intelligent cooperative behaviors will pave the way for achieving realistic flexible automation systems. In specific, the research conducted through the VERAM (Virtual Environments for Robotics, Automation and Manufacturing) at the Software Automation and Intelligence Laboratory (SAIL) at Tennessee Technological University, investigates the applicability of agent-based design techniques in the development of interactive software components designed as ICE (Intelligent Coordinating Entities)elements.1. IntroductionMost modern day, industrial-strength systems are complex and operate in a dynamically changing environment. Building software systems that are easier to use in such applications implies addressing the issues of increasing complexities in managing processes and communication between processes .Current system simulation and prototyping techniques, based on highly simplistic models,involve a costly, labor-intensive modeling and development process. In order to manage complexity,it is very important to have tools, notations and methodologies that support the designer’s work during design resolution. Moreover, advances in virtual reality (VR) have made it feasible to directly utilize VR for the modeling and simulation in virtual environments. The use of virtual reality in simulating manufacturing control software environments give designers the opportunity to play a pro-active role in identifying flaws and optimizing the design. Current techniques do not allow the designer to be progressively involved in the design process. Often,the designer is limited to the initial specification stages and her/his experience is not exploited throughout the design process.Software entities, called Intelligent Coordinating Entities - or ICE, as proposed by this research, go a step further than intelligent agents. ICE elements extend the agent-oriented design approach to implicitly include a coordination framework implemented through different enabling technologies (such as DCOM/CORBA/JINI), which are necessary to bring such intelligent coordinated behaviors to fruition. In a system, an ICE element (ICICLE),therefore, not only consists of a representation,decision and reaction structure, but also includes the mechanisms to orchestrate such reactions in a coordinated fashion using a well-defined communication structure as the defining baseline for all ICICLES. Thus, a network-centric communication structure forms the core of every ICE element (similar to the hexagonal structure in the snow crystal analogy). Depending on the enabling technology to be used, it may or may not be necessary to include the code for the communication structure within each ICICLE. However, depending upon the other players in the coordination, an ICICLE can selectively choose the appropriate information it is willing to share with the group by means of a dynamic model of its currently available services.2. ICE Element DesignA flexible communication structure forms the basis of any ICE element. A software system designed withICE elements, or ICICLES, assumes a networked structure with dynamically interacting ICECLES. The interaction capability of an ICECLE is time-variant and hence may be different at different time instances depending upon the composition of the system and the other elements involved in the interaction. Each ICECLE may or may not have a corresponding hardware resource associated with it -the assumption being that all hardware devices come with an inherent ability to communicate with other existing group elements. This communication capability is platform and language independent and hence can be brought forth by using any standard specifications such as CORBA [1], JINI[2], etc.On top of the above communication framework, coordination capability is built by maintaining a dynamic "information-sharing" model within each element. This model is built / updated and used to verify the availability of a service / information by the requested ICICLE (server), when the service is requested by another ICICLE (client). This model is dynamically generated / updated at the server by exploding / imploding details on specific services depending on the interaction client. With the server providing such a service, the client will be able to assimilate information about the server and hence act either pro-actively or reactively within the software system.For example, assume that this technique has been used to build a automated factory floor. For simplicity, assume that each factory floor resource, would therefore, in essence be an ICICLE. Assume that a resource, say R1, at the beginning of a job flow is stuck with a difficult job. With its communication structure, it is able to communicate this information to other members of the group if it chooses to do so. Correspondingly, these other members can reevaluate their goals / strategies by spending their resources more productively by reacting intelligently to this information. However, in designing such a system, the first order of priority is to devise a mechanism to discern the information to be shared! To accomplish this we will use a Petri-net based technique to extract all the decision-points in the model. The hierarchical Petri-net model with the capability for selective expansion / contraction of subnets allow for selective information sharing.ICICLES are based on a flexible communication structure with incrementally built flexible coordination, reasoning and other application-specific modules built on domain-specific needs. Thus,software systems can be built as a combination of components that vary in their capabilities ranging from ones that are highly intelligent to those that are dumb. Moreover, the dynamic cooperation model allows for components that can have a time-dependent interaction behavior. In a system with multitude of ICICLES, the presence or absence of a particular component would thus be of little significance as long as the system is build with redundant structures- if a component is present and is willing to cooperate it will be utilized. This can allow for the flexibility to leverage industry and vendor partnerships.3. Current Projects at SAIL3.1.Self Organizing Software EntitiesThis project attempts to design simple software structures in JAVA using agent-based design concepts and KQML messages to build systems with interacting and coordinating structures. The project is focused on developing a prototype system consisting of different-sized boxes that cooperate to execute a command (or goal) - i.e. organizing into a given pattern such as a line, circle, etc. The experience gained from implementing this prototype is to be used to implement a flexible system consisting of group of machines with differing capabilities that can coordinate / cooperate among themselves to design and execute dynamic process schedules for a collection of jobs (or goals). A screen shot of theabove project is shown in Figure 1.Figure 1. Self Organizing Structures3.2. Home AutomationA Simple home automation scheduler has also been developed. It involves the virtual simulation of two simple home appliances - a coffeepot and a light that are programmed through a graphical interface. These appliances can be set to go off or on either at random or scheduled (either periodically or as a one-time operation at some specified time). A snapshot of this application is shown in Figure 2 - note the subtle shade difference in Figure 2b due to the light ON state. The application has integrated sound capabilities wherein the sound of the coffeepot inoperation diminishes as we move away from the appliance within the virtual world.3.3. Naval Radar SimulationA two-level distributed simulation environment for interference detection and frequency assignment in naval radar units has been developed. The Java-based simulation is designed as a two level architecture incorporating i) multiple radars on each ship controlled through a lower-level intra-ship interference control module and, ii) multiple ships in a group coordinated through ahigher-levela. OFF Stateb. ON State (auto-timer driven)Figure 2. A Simple Home Automation PrototypeFigure 3. Lower-Level Control Agent Interfaceinterference control module. Each ship’s interference detection and control mechanism is composed of these two separate levels, called a control agent, dynamically coordinating with one another in eliminating interference problems. In addition, the control agent is a multi-threaded architecture that incorporates the maintenance of a distributed data base system that contains periodically updated radar information. Current implementation involves centralized, locally autonomous and negotiation-based interference resolution strategies. A screen shot of the interference resolution process is shown in Figure 3.3.4.Power Plant Distribution VisualizationO perator-visualization of observable data in a power system control center will enable efficient and reliable decision-making, especially under stressful conditions. Open access environments and the ever-increasing crunch on generation capacity augmentation are creating unprecedented demand for the availability of more "processed" information to help in the selection of safe and economically acceptable options. Recent advances in computing technology have forced data presentation to a power system control center operator, to evolve from simple one-line diagrams to animations and 3-D environments. This recently begun research project will strive to develop a visualization package for a power system control center. The prototype will focus on functions such as: available capacities of generation and transmission, voltage levels, and some financial cost information. Data for this prototype can be gathered from individual databases or from analysis programs making on-line calculations. The underlying philosophy will be to minimize the use of costly and generalized software packages, and to incorporate ergonomic and human-centric design principles. The project will exploit advances in virtual reality and multimedia techniques for interactive design, analysis and simulation. Such virtual environments offer the ability to evaluate current capabilities, future plans and schedules, without the investment of vast technical and financial resources. As an experimentation tool, the interactive virtual environment will allow power companies to conduct a rapid and thorough assessment of risks, affordability, ease of production, and other impacts on power generation capabilities. Moreover, simulation in the virtual environment will provide a foundation for continued experimentation and process development, thereby aiding market advancement. Finally, simulated training in such virtual environments promises to be a useful approach to provide operators the necessary competitive advantage. The prototype software will focus on the human-centric design principles to aid the operator in navigating through complex virtual environments. Specifically the capabilities will include:1.Rapid Navigation Capability to ReduceNavigational "Stress" on the Operator: The package will provide the capability for rapid navigation through three-or-four menu or button-controlled customized views (top, bottom, side-view, etc.) of the control center data that would ease the navigational "stress" on the operator. In each such view, the operator will have the capability to select grid lines both individually, or as a group and the ability to zoom to specific locations. In such a virtual environment with rapid navigational facilities, the operator has the obvious advantage of having greater time to concentrate on issues that boost productivity and not spend their time on issues that induce stress and fatigue.2.Exploit Language-Specific PerformanceCapabilities: The software will be multi-language based - it will derive from the capabilities of the Virtual Reality Modeling Language (or VRML) to create and visualize a virtual environment, while it will use the logic and processing capabilities of Java to interact with the VRML created virtual environment. The advantage to this approach is that both VRML and Java are "good" at what they do and do it well with respect to performance considerations.3.Seamless Environment to Reduce HardwareCapability Restrictions: Both VRML and Java are neither platform-specific nor location-specific languages, and hence can provide a seamless virtual environment without specific hardware/software demands. The environment can be implemented on a multitude of machines and execute satisfactorily on quite a few widely used operating systems. It can be built with the ability to read VRML data, thereby allowing thesharing of virtual worlds created by other software packages.4. Licensing and Other Indirect Costs: Finally,such a virtual environment will be very cost-effective in terms of hardware and software acquisition, as well as their maintenance. A network of desktop machines with mostly public domain software can host a huge database of generation and distribution data. Also, the machines can be serviced much more cheaply through local information technology departments. Moreover, Java will provide the necessary interoperability through Java Database Connectivity (JDBC) calls to any database that holds generation and distribution data. Due to its network-centric development, it provides better features for local area network operations and database connectivity.4. Expected Research ImpactThe ICE based approach will lead to the development of a comprehensive design framework for developing flexible, interactive, distributed software systems, with domain-specific intelligence. The research will provide a new thrust and direction for the development of control software systems. A potential payoff will be the efficient and flexible control strategies that may be easily derived from the design efforts. Although the advent of software based control has greatly improved our manufacturing sectors, much of our current day woes can also be easily attributed to this inherent inflexibility provided by software based control. However, this problem is not solely restricted to this domain. Often, the problem is because the software is not effectively designed and tested. A simple by-product of the research will be a solution to identifying necessary test conditions early in the design. Thus, the proposed research may produce results, which are widely applicable in various engineering and computer science areas.5. ConclusionsIn addition to the above stated projects that have been directly impacted by the use of the equipment purchased through this NSF equipment grant, the equipment has also provided the necessary computational power for the following projects to students on campus working towards their Master’s / Ph.D. thesis in engineering fields. Ms. Eik Lang Lau from the chemical engineering program at Tennessee Technological University has used the systems for reaction modeling wherein she has used the machine for studying reactions between calcium hydroxide and fly ash using cellular automaton techniques.6. AcknowledgementsThe above research projects were supported, in part, by a research equipment grant from the Division of Design and Manufacturing (DMI 9896235).7. References1.Object Management Group, "Common ObjectRequest Broker Architecture ", 2.Jim Waldo, "The JINI Architecture for Network-Centric Computing", Communications of the ACM, July 1999. pp.77-82.。
QoS , , .
ISSN 1000-9825, CODEN RUXUEW E-mail: jos@Journal of Software, Vol.17, No.6, June 2006, pp.1381−1390 DOI: 10.1360/jos171381 Tel/Fax: +86-10-62562563© 2006 by Journal of Softwar e. All rights reserved.QoS获益驱动的中间件调度框架研究∗张文博1,2,3+, 陈宁江4, 魏峻1, 黄涛1,21(中国科学院软件研究所软件工程技术研发中心,北京 100080)2(中国科学院软件研究所计算机科学重点实验室,北京 100080)3(中国科学院研究生院,北京 100049)4(广西大学计算机与电子信息学院,广西南宁 530004)A QoS Benefit Driven Scheduling Framework for MiddlewareZHANG Wen-Bo1,2,3,+, CHEN Ning-Jiang4, WEI Jun1, HUANG Tao1,21(Technology Center of Software Engineering, Institute of Software, The Chinese Academy of Sciences, Beijing 100080, China)2(Key Laboratory of Computer Science, Institute of Software, The Chinese Academy of Sciences, Beijing 100080, China)3(Graduate School, The Chinese Academy of Sciences, Beijing 100049, China)4(College of Computer, Electronic, and Information, Guangxi University, Nanning 530004, China)+ Corresponding author: Phn: +86-10-62630989 ext 205, Fax: +86-10-62562538, E-mail: wellday@Zhang WB, Chen NJ, Wei J, Huang T. A QoS benefit driven scheduling framework for middleware. Journalof Software, 2006,17(6):1381−1390. /1000-9825/17/1381.htmAbstract: The first-come first-served scheduling framework adopted by most application servers has been provedto be inappropriate for dealing with unexpected overload for Internet-oriented Web applications. Considering thedeficiency of the existing work from the architecture view, this paper presents a scheduling framework based on thenotion of QoS benefits, which contains several cooperating components to guarantee the QoS requirements of theapplications. The QoS benefits used to evaluate the QoS guarantee provided by the server according to the QoSrequirements of the applications and the resource management based on the QoS benefits will help to provide abetter QoS guarantee. Experimental results confirm the efforts on the OnceAS application Server.Key words:Web application server; QoS benefit; scheduling framework; resource management摘要: Web应用服务器目前普遍采用的先到先得式(FCFS)的调度框架在过载时难以保障应用的服务质量(QoS)需求.QoS获益驱动(QBD)的调度框架是一种针对这些不足而提出的请求调度解决方案.QoS获益根据应用的QoS需求得到,用于评价QoS保障对应用需求的满足情况.QBD调度框架包含了多个用于保障应用QoS需求的组件,实现了基于QoS获益的资源规划算法,能够提高服务器对应用QoS需求的保障能力.在OnceAS平台上的实验结果验证了QBD调度框架的有效性.关键词: Web应用服务器;QoS获益;调度框架;资源管理中图法分类号: TP393文献标识码: A∗ Supported by the National Natural Science Foundation of China under Grant No.60573126 (国家自然科学基金); the National GrandFundamental Research 973 Program of China under Grant No.2002CB312005 (国家重点基础研究发展规划(973))Received 2006-01-24; Accepted 2006-03-131382 Journal of Software软件学报 V ol.17, No.6, June 2006面向Internet的Web应用越来越关注提供给用户的服务质量(QoS),比如信息门户和电子商务应用等,难以容忍的响应时间和服务失败将给应用带来损失.由于Internet用户数量不可预知,这些应用经常会遭遇突发性的巨大的访问量,比如出现热点新闻或是节日集中购物的时候.由于用户的负载超过服务器的处理能力,所以这种情况称为过载(overload)[1].此时,应用需要服务器按照其QoS需求提供有区别的保障,使其可以在过载的时候继续为部分重要的请求提供服务.Web应用服务器是Web应用的主要支撑平台,现有的应用服务器的调度框架由于在体系结构方面缺乏对QoS保障的考虑,因此只能采取先到先服务(first-come fist-serve,简称FCFS)的调度策略.该调度框架不具备QoS 保障能力,过载时,大量请求不能得到及时处理,导致后到的重要请求不能得到及时的响应,几乎所有的用户都需要经历难以忍受的响应时间,而响应时间是影响Web应用用户QoS感受的主要因素.因此,Web应用服务器需要一种新型的调度框架为Web应用提供过载时的QoS保障.通过请求调度缓解服务器的过载是中间件领域的研究热点之一,已有的研究成果提出了一些解决服务器过载的方案,包括根据负载情况[2]或是用户优先级[3−5]进行准入控制等,但这些方案都不能完全满足Web应用复杂、灵活的QoS需求.主要的不足体现在:1) 缺乏在体系结构上对应用QoS需求的保障,由于中间件提供的QoS 保障涉及对应用QoS需求的管理、请求调度以及资源优化等多个环节,因此必须在体系结构上完善服务器的调度框架;2) 缺乏对应用QoS需求的保障能力,如对于不同类型的请求其QoS需求可能并不相同,对应用的影响也不相同,服务器提供的QoS保障需要充分考虑这种需求;3) 缺乏针对应用QoS需求的资源管理能力,当服务器过载时,不能根据应用QoS需求来优化资源分配.针对传统Web应用服务器调度框架及现有研究成果的不足,本文从中间件体系结构方面研究了请求调度与应用QoS需求管理以及资源管理的关联,并在此基础上研究了一种QoS获益驱动的请求调度框架.该框架通过对应用的QoS需求进行管理,实现了有差别的QoS保障以及对QoS保障效果的应用评价(即QoS获益);在对运行环境监控与评估的基础上,建立QoS获益与资源之间的映射关系;通过QoS获益驱动的资源管理,实现了对应用QoS保障效果的优化.实验结果表明,该调度效果可以提高应用服务器对于应用QoS需求的保障能力.本文第1节介绍中间件的调度框架应如何提供对应用QoS需求的保障.第2节介绍QoS获益驱动的调度框架的主要原理与关键技术.第3节通过实验来验证该调度框架对应用QoS需求的保障能力.第4节进行相关工作的比较.最后是总结以及对未来工作的展望.1 应用QoS保障的需求目前应用服务器普遍采用的是一种FCFS调度框架,如图1所示.这种调度框架在体系结构上缺乏对应用QoS需求的考虑,只有单一的请求队列,调度资源只能提供一种先到先得式的管理策略,因此无法实现请求的区分以及根据QoS需求管理资源.FCFS调度公平性较好,先到达的请求总是会优先得到服务.然而当服务器过载时,由于服务器处理能力不足造成大量请求不能得到及时处理,关键的请求也必须在请求队列中长时间等待,或是因为不能进入请求队列而导致服务失败.因此,应用在服务器过载时对所有用户表现出一种“无响应”状态.而响应时间是Web应用用户最关注的QoS指标,因此,FCFS的调度框架不具备应对服务器过载的能力.为了在过载时保障应用的QoS需求,应用服务器需要一种新型的调度框架,它应该具有如下能力:•管理应用的QoS需求.不同的应用请求具有不同的QoS需求[6],调度框架需要有能力管理这些需求,包括为应用提供QoS需求的描述方式以及服务器对QoS需求描述的解析能力.•请求区分.当服务器过载时,由于处理能力不足,所以调度框架必须能为具有不同QoS需求的请求提供有差别的服务,如优先处理一些重要的、需要及时得到响应的请求.• QoS保障效果的评价.调度框架必须有能力评价请求的调度对应用QoS需求的保障效果,并根据这种评价提高对应用QoS需求的满足能力.•资源优化.服务器过载时,系统资源处于一种高度紧张的状态,因此,服务器必须有能力根据应用的QoS需求和系统的运行状况优化资源的分配,以取得更好的QoS保障效果.张文博 等:QoS 获益驱动的中间件调度框架研究 1383下一节将根据以上的需求扩展现有应用服务器的调度框架,使其具备过载时对应用QoS 需求的保障能力.EJB containerWeb container S c h e d u l i n g r e s o u r c eRequest queueManagement kermelRequests arrivingRequests processingFCFS schedulingServer 2Server 3Server 1Web application serverFig.1 FCFS scheduling framework图1 FCFS 调度框架2 QoS 获益驱动的调度框架及其关键技术根据前面的分析,体系结构上的不足是目前请求调度框架难以提供QoS 保障的关键.因此,为了对原有的FCFS 调度框架在体系结构上进行扩展,针对前面提到的保障应用QoS 需求需要具备的能力,我们分别提出了相应的功能模块,包括QoS 需求管理、请求管理、QoS 获益评估以及资源管理,每个功能模块作为一个系统服务启动并接受管理内核的管理.QoS 获益驱动的调度框架原理如图2所示..... . . Resource unitRequests managementR e s o u r c e p l a n n e rR e q u e s t s c l a s s i f i e rQoS-Benefit driven schedulingWeb application serverServer nServer 2 Server 1Management kermelEJB containerWebcontainer Requestsarriving Requests processingR e s o u r c e m a n a g e m e n tRequest queueQoS requirementsinformationQoS evaluator QoS benefit evaluatorQoS requirements deployerQoS requirements management Fig.2 QoS benefit driven scheduling framework图2 QoS 获益驱动的请求调度框架该调度框架的工作原理是:应用部署过程中,QoS 需求部署器解析应用的QoS 需求,并生成相关的QoS 需求信息;运行过程中,请求管理模块根据QoS 需求信息将请求分类,并放入相应的请求队列;QoS 获益评估模块根据应用的QoS 需求信息对中间件提供的QoS 保障状况进行评价,并得到评价结果——QoS 获益;服务器线程等调度资源是影响服务器性能的关键资源,资源管理模块将根据QoS 获益情况对调度资源进行优化管理,以优化应用整体的QoS 获益.下文将分别介绍调度框架中主要模块的作用及其关键技术.1384Journal of Software 软件学报 V ol.17, No.6, June 20062.1 QoS 需求管理2.1.1 应用的QoS 需求分析与描述服务器的吞吐量和用户得到的响应时间是应用关注的主要性能指标[8],服务器必须同时权衡这两个指标以提供更好的QoS 保障,即在保证用户可以接受的响应时间的同时,为更多的用户提供服务.应用对不同的请求得到的响应时间满意程度可能并不相同[6],本文使用下面的方式来描述应用对请求的QoS 需求:• Texpected:称为请求的期望响应时间.当请求的响应时间低于Texpected 时,则认为用户得到了满意的QoS;• Ttimeout:称为请求的超时时间.当请求的响应时间超过Ttimeout 时,服务器为该用户提供的QoS 将难以被接受,对应用也不再有意义.而当请求的响应时间位于Texpected 和Ttimeout 之间时,用户会感受到服务质量降低,但仍然会选择接受服务.T expected 和T timeout 描述了应用对请求的响应时间方面的QoS 需求.此外,不同类型的请求对应用可能有不同的影响,比如位于电子交易后期的订单请求要比前期的浏览请求更重要.因此,本文使用以下属性来表达请求对应用整体QoS 的影响:• BScale:称为请求的QoS 获益系数,用来描述请求对于应用的重要程度,反映了该请求对于应用得到的整体QoS 的影响.BScale ,T expected ,T timeout 这3个属性构成了请求的QoS 需求属性,应用可以通过为请求定义这些属性来表达QoS 需求.2.1.2 QoS 需求的部署为了维护应用在应用服务器平台上的可移植性,本文基于部署过程来实现请求的QoS 需求从描述到运行时信息的映射.应用通过部署描述文件来描述QoS 需求,即使用BenefitSchema.xml 文件来声明应用各类请求的获益属性,其DTD 的格式如图3所示.MethodName :请求所关联的业务方法名称,请求管理模块将根据该信息对请求进行分类处理;BScale :请求的QoS 获益系数,取值为正整数; Fig.3 Benefit schema descriptor图3 获益方案描述符TtimeoutTexpected BScale MethodNameBenefit*Texpected :请求期望响应时间,单位为ms; Ttimeout :请求超时时间,单位为ms.QoS 需求部署器将部署描述信息形成运行时的信息,主要是将请求的QoS 需求属性与其业务方法机进行关联.服务器在部署的过程中,将根据请求的获益属性建立获益类别及相应的请求队列.请求管理和QoS 获益评估模块将使用到部署后的QoS 需求.2.2 请求管理FCFS 调度框架中的单一请求队列不具备区分请求的能力.为了能够应对服务器的过载,服务器必须有能力对请求进行区分,并根据应用的QoS 需求对不同的请求实现有差别的处理.因此,ABDS 框架中的请求管理模块扩展了FCFS 单一的请求队列,采用请求分类器加多个请求队列的结构.请求管理模块根据请求的QoS 需求进行分类,由于具有相同QoS 需求的请求的QoS 获益属性相同,我们把具有相同QoS 属性的请求称为一个QoS 获益类别,同属一个QoS 获益类别(QoS benefit category,简称QBC)的请求共用一个请求队列.定义1. QoS 获益类别.QBC=(BScale ,T expected ,T timeout ),用于标识一组具有相同QoS 需求的请求.请求分类器根据请求的获益属性将请求放置到相应的请求队列,每个请求队列采取先进先出的调度方式.使用请求的获益属性作为分类标准具有如下特点:1) 便于对获益类别的请求进行整体的评估,可以进行针对QoS 需求的资源优化管理;2) 每个请求队列采取先进先出的调度,该分类方式可以保障具有相同QoS 属性的请求得到公平的处理.张文博 等:QoS 获益驱动的中间件调度框架研究 13852.3 QoS 获益评估2.3.1 请求的QoS 获益评估服务器每完成应用的一个请求,都可以看作应用得到了一定程度的获益,获益的大小反映了应用对服务器提供的QoS 保障的评价,我们将其称为QoS 获益.该获益的大小取决于两个因素:一是根据请求的QoS 需求对其完成情况(即响应时间)的评价;二是请求的QoS 获益系数大小.本文使用一个评价函数来计算应用对请求完成情况的评价,即若请求q 的响应时间为t ,则q E =f E (t )称为请求完成质量,其中f E (t )∈[0,1]称为请求评价函数.本文使用的请求评价函数f E 的函数关系如图4所示.当t 不超过T expected 时,应用认为服务器按质量完成了请求q ,q E 为1;当t 超过T timeout 时,应用认为服务器处理q 失败,q E 为0;而当t 位于T expected 和T expected 之间时,q E 随着t 的增加从1到0递减.因此,请求完成质量反映了应用对服务器处理q 的满意程度.定义2. 请求QoS 获益.若请求q 的请求完成质量为q E ,QoS 获益系数为BScale ,则q B =BScale ×q E 称为q 的QoS 获益.定义3. 应用QoS 获益.若服务器在时间T 内完成应用A 的请求集合为{q 1,q 2,…,q n },每个请求q i 的请求QoS 获益为,则称为应用A 在T 时间内的应用QoS 获益,简称应用QoS 获益.i B q ∑=ni B i q 1q B =f B (t )=q E t01(t >T timeout )(t ≤T excepted )(Ttimeout −T excepted ) (T excepted <t ≤T timeout −t )(T timeout −t )/1T excepted T timeout 应用QoS 获益体现了应用对服务器所提供QoS 的一种综合评价.如果在等量的时间T 1和T 2内,应用A 的QoS 获益分别为B 1和B 2,且B 2>B 1,则认为服务器在T 2时间内提供的QoS 保障要优于T 1. Fig.4 Evaluation function of requests图4 请求评价函数2.3.2 预期QoS 获益对于同一QoS 获益类别的请求q ,其请求QoS 获益取决于响应时间.根据该获益类别请求的到达和处理情况,可以预测请求的响应时间并计算相应的请求QoS 获益,进而可以预测该获益类别的QoS 获益.通过这种方法得到的QoS 获益称为预期QoS 获益.由于负载和服务器运行环境的动态性,QBD 调度框架中的获益评估器周期性地计算获益类别的预期QoS 获益.对于某获益类别QBC i ,其调度资源数量为r .在每一个采样周期T (即获益评估器进行计算和评估的周期)中,前一采样周期的请求到达率和请求实际处理时间已知,则T 周期内属于QBC i 的任意请求q 的预期响应时间只依赖于r ,即若请求到达和处理稳定时, QBC i 的预期QoS 获益只取决于该获益类别拥有资源的情况,我们使用B ′＝f EQB (r )表示,容易确定B ′随r 增加非减.同理,我们可以预测当QBC i 增加部分资源∆r 后的预期QoS 获益+′B =f EQB (r +∆r )和减少资源∆r 后的预期QoS 获益−′B =f EQB (r −∆r ).我们定义win B ′(∆r )=−B ′和(∆r )＝B ′−为QBC +′B loss B ′−′B i 在∆r 资源约束下的预期QoS 获益增益与预期QoS 获益损失.通过这种方式,在服务器资源与获益类别的QoS 获益之间建立起一种关联,调度框架将根据这种关联来进行QoS 获益驱动的资源管理,提高应用的整体QoS 获益. 2.4 资源管理 2.4.1 资源规划FCFS 调度框架中并不具备根据应用QoS 需求进行资源管理的能力.当服务器过载时,资源分配不能优先保障关键请求的处理,所有的请求都需要长时间地等待处理,影响了服务器的性能.本文的QoS 获益驱动请求调度框架中对资源按照请求的获益类别进行管理,为每个获益类别的请求队列分配独立的资源单元,每个资源单元包含一定量的调度资源,这些调度资源仅处理对应请求队列中的请求.服务器通过调整各获益类别的资源单元的大小,实现对不同类别的请求有差别的QoS 保障.这种为不同获益类别请求分配和调整资源的行为称为资源规划.定义4. 假设服务器提供给应用的资源总量为R ,应用共有n 种获益类别QBC 1,QBC 2,…,QBC n ,它们的资源1386 Journal of Software 软件学报 V ol.17, No.6, June 2006单元的大小分别为r 1,r 2,…,r n ,则{r 1,r 2,…,r n }称为一个资源规划,其中=R .∑=ni i r 1资源规划器根据每个获益类别QoS 获益情况进行资源规划,寻求应用整体QoS 获益的提高. 2.4.2 QoS 获益驱动的资源管理服务器对资源进行管理的主要目的是实现应用QoS 获益的优化.为了提高资源管理的效率,我们根据获益类别的预期QoS 获益进行资源规划.由于运行环境的变化,资源规划需要周期性地进行.以下给出了QoS 获益驱动的资源规划算法.在算法中,设定一个最小的资源约束∆r min .假设应用共有n 个获益类别,当前的资源规划为{r 1,r 2,…,r n }.算法. QoS 获益驱动的资源优化. 1.每个调度周期T 之前执行:2. 1) 对每个获益类别QBC i :执行:3. a) 若其资源r i ≥∆r ,则计算其在∆r 资源约束下的预期获益增益(winB ′(∆r ))i 与预期获益损失(lossB ′(∆r ))i ; 4. b) 否则计算其在∆r 资源约束下的预期获益增益(winB ′(∆r ))i ;(loss B ′(∆r ))i =0; 5. 2) 如果存在两个获益类别QBC i 和QBC j ,并满足条件(B ′(∆r ))win i >(B loss ′(∆r ))i 且(loss B ′(∆r ))i ≠0且二者之差最大,则: 6. a) 制定新的资源规划{r 1,…,r i+∆r ,…,r j −∆r ,…,r n }; 7.b) 实施该资源规划;8. 3) 否则,∆r =(∆r −∆r min )>0?(∆r −∆r min ):∆r min ; 9. 4) 等待下一调度周期开始.由于每次优化仅涉及两个获益类别QBC i 和QBC j 的∆r 资源变化,而它们的应用整体QoS 获益的影响分别为((∆r ))winB ′i 和((∆r ))loss B ′i ,条件(win B ′(∆r ))i >((∆r ))loss B ′i 将保证∆r 资源从获益类别j 转移到获益类别i 可以使应 用整体的预期QoS 获益增大.仅选择预期QoS 获益变化最大的两个获益类别进行资源规划的原因在于:防止由于环境运行变化导致频繁地进行资源规划,而只是通过局部资源调整使应用的QoS 获益处于一种较优的状态.3 实验数据及分析本文讲述的QoS 获益驱动的请求调度框架已经在应用服务器OnceAS 原型系统中实现,本节将通过实验来验证其对应用QoS 需求的保障效果. 3.1 实验设计实验用例为一个在线购书系统.为简单起见,我们只关注如下几类主要的操作:浏览(browse)、选购(addCart)和购买(purchase).用户的测试流程设计如下:• 一半用户依次提交浏览、选购和购买请求; • 一半用户在浏览后直接结束测试过程;• 用户的每个请求在超时后不再提交下一个请求,而是选择结束.该应用的用户请求是以会话(session)组织的.从应用的获益来看,只有最后购买的用户会话才可能为应用产生真正的利益.没有购买意愿的会话和由于QoS 保障原因(请求超时)而结束的会话都不会为应用带来实际的利益.测试程序会统计测试过程中服务器完成会话的情况来反映对应用利益的保障.实验在分布式环境中进行,包括应用服务器、数据库服务器和客户机3个机器节点,其部署如图5所示. 在本实验中,调度框架中资源是指服务器的处理线程,总数为100,最小资源约束为∆r min =2,调整周期为T =10s.测试时间分为初始时间T up =60s 、测试时间T steady =300s 和结束时间T down =30s.测试机在T up 和T down 只发送请求,不记录测试结果;只在T steady 内统计请求处理结果.张文博等:QoS获益驱动的中间件调度框架研究1387Fig.5 Deployment environment of test图5 实验部署环境3.2 请求QoS需求分析从应用的需求分析可知,3类请求由于功能不同,所以对应用的影响也不同,它们对于应用的重要程度分别为:浏览并不会直接产生实际的应用获益,所以重要性最低;选购会产生潜在的应用获益,重要性其次;而购买会产生直接的应用获益,重要性最高.根据上面对各类请求重要程度的分析,其BScale和T timeout按照请求的重要程度依次增加,而请求的期望时间则需要与用户的期望值相吻合,比如浏览和购物属于需要涉及数据库访问,其T expected较高;而选购通常不涉及数据库访问,所以其T expected较低.各类请求的获益属性见表1.Table 1Benefit attributes of different requests表1不同请求的获益属性声明T expected (ms) T timeout (ms)BScaleBrowse 1 2000 2000AddCart 2 1000 3000Purchase 4 2000 60003.3 测试结果与分析在本节的数据图表中,分别使用FCFS和QBD表示原有的采用FCFS调度框架和采用QoS获益驱动调度框架的OnceAS服务器.测试条件分为正常和过载两种,分别使用underload和overload表示这两种测试条件.图6(a)~图6(c)分别表示OnceAS服务器在采用不同调度框架后,在正常和过载情况下的平均相应时间、吞吐率和对应用会话的完成情况.从图6(a)可以看出:当服务器过载时,采用QoS获益驱动的调度框架后,各类请求的平均响应时间都稳定在请求的期望响应时间附近,而采用FCFS调度框架时,各类请求的平均响应时间都大幅度地超过了请求的超时时间.这说明QBD调度框架可以有效保障应用的QoS需求;图6(b)表明在服务器过载时,与FCFS调度框架相比,QBD调度框架完成了更多的重要请求,实现了对不同请求有差别的QoS保障;图6(c)表明QBD调度框架更好地体现了应用的利益,与FCFS调度框架相比,用户会话的完成率提高了349% (122.2/27.2).由于调度框架按照获益类别的预期获益情况进行管理,因此能够产生较高QoS获益的请求将得到更多的调度资源.因此在QBD调度框架中,获益系数较高的请求平均响应时间较短(与请求的期望值接近),出现超时的可能性小,因而吞吐率也较大.同理可以解释图6(c)中会话完成率的提高,由于处理比较重要的选购和购买请求QoS获益较大,QBD调度框架对其QoS保障效果较好,因而整个会话完成的比例较高.图7表示了在服务器过载时的获益率变化情况.可以看到,QBD调度框架使用预期QoS获益可以更有效地调整服务器的资源规划,从而获得更多的应用获益.服务器通过对预期获益的评估,可以使对资源的优化更有目的性,从而使服务器处于一种较优的状态.1388 Journal of Software 软件学报 V ol.17, No.6, June 2006Average response time under different conditions Throughput rate under different conditions A v e r a g e r e s p o n s e t i m e (m s )T h r o u g h p u t r a t e (r e q s /s )76543210QBD-underload FCFS-underload QBD-overload FCFS-overloadBrowse addCart purchase (a)(b)Session rate under different conditions S e s s i o n r a t e (s e s s i o n s /s )140120100806040200QBD-underload FCFS-underload QBD-overload FCFS-overload(c)Fig.6 Test results under different conditions图6 不同条件下的性能测试结果Benefit rate fluctuation120010008006004002000B e n e f i t r a t e (B /s )Test time (s)Fig.7 Benefit rate fluctuation during test 图7 测试过程中的获益率变化曲线图8则统计了在服务器过载时的吞吐率和会话完成率.可以看出,QBD 调度框架没有一味追求更高的系统吞吐率,而是根据应用的QoS 需求提供相应的保障.系统的吞吐率与FCFS 相比虽然略低,但是对应用会话的支持则明显增加,更好地体现了应用的利益.FCFS 调度框架对请求不作区分,过载时完成了较多的浏览请求,而QBD 图8 测试数据变化曲线张文博等:QoS获益驱动的中间件调度框架研究1389综合以上实验结果,QBD调度框架与FCFS调度框架相比,可以针对应用的QoS需求进行QoS保障,从而更好地保障应用的利益.4 相关工作比较服务器过载是面向Internet的Web应用经常面临的问题,如何保障服务器在过载时提供应用所需的QoS 保障是目前研究的热点之一[10].通过动态增加服务能力可以提供服务器过载时的QoS保障,然而为峰值负载准备计算资源并不是一种十分可行的解决方案.在服务器过载时,通过准入控制来优先选择重要的请求进行处理,是较为理想的解决方案.如文献[3−5]针对用户优先级进行请求分类,较好地解决了不同优先级用户的服务差分问题.然而,这种解决方案并不能更细粒度地区分同优先级用户的不同请求.已有研究成果[6,7]表明,不同类型的请求的QoS需求并不完全相同,如果对这些请求不加区分地处理,可能会给应用的利益带来损害.本文与上述相关工作相比,主要有以下几个特点:1) 从体系结构的角度研究了提供面向应用需求的QoS保障所需的功能组件及其相互关系;2) 使用反映应用需求的QoS获益作为提供QoS保障的评价,增强了对应用QoS需求的保障能力;3) QoS获益驱动的资源管理提高了服务器根据需求使用资源的能力.另外,服务器的自优化也是提高服务器QoS保障能力的研究内容之一,如文献[9]采用反馈控制理论对Web 服务器的静态页面请求进行自优化,但该方法在以动态内容为主的复杂Web场景中的应用还比较困难;文献[11]采用排队论的方式对Web系统的请求进行优化;文献[2]则采用短任务优先的调度方法,提高系统的吞吐率.本文与上述工作以服务器的性能为优化目标不同,它是以根据应用需求得到QoS获益作为服务器优化的目标,这在根本上体现了应用的性能需求,同时,调度框架使用预测技术来进行资源管理的优化,该方法有助于提高资源的优化效率.5 结论与工作展望本文针对Web应用服务器现有调度框架在体系结构上对应用QoS需求保障能力的不足,提出了一种QoS 获益驱动的调度框架.该框架引入了多种用于保障应用QoS需求的设施:QoS需求管理提高了服务器对应用QoS需求的理解能力;请求管理使服务器具备了按照应用QoS需求提供QoS差分的能力;QoS获益评价及其驱动的资源管理则使服务器具备了根据应用QoS需求进行自我优化的能力.实验结果表明,该框架可以有效地提高应用服务器过载时对应用QoS需求的保障能力.未来的工作将进一步研究如何校正应用QoS需求与应用利益的关系,以及如何在更大范围内进行资源优化的问题.References:[1] Crovella M, Bestavros A. Self-Similarity in World Wide Web traffic: Evidence and possible causes. In: Reed DA, Gaither BD, eds.Proc. of the ’96 ACM SIGMETRICS Int’l Conf. on Measurement and Modeling of Computer Systems. New York: ACM Press, 1996. 160−169.[2] Elnikety S, Nahum E, Tracey J, Zwaenepoel W. A method for transparent admission control and request scheduling in e-commerceweb sites. In: Fledman SI, ed. Proc. of the Int’l WWW Conf. New York: ACM Press. 276−286.[3] Urgaonkar B, Shenoy P. Cataclysm: Handling extreme overloads in Internet applications. In: Chaudhuri S, Kutten S, eds. Proc. ofthe 23rd Annual ACM Symp. on Principles of Distributed Computing. New York: ACM Press. 2004. 390−390.[4] Schroeder B, Harchol-Balter M. Web servers under overload: How scheduling can help. ACM Trans. on Internet Technology(TOIT), 2006,6(1):20−52.[5] Bhatti N, Friedrich R. Web server support for tiered services. IEEE Network, 1999,13(5):64−71.[6] Bouch A, Kuchinsky A, Bhatti N. Quality is in the eye of the beholder: Meeting users’ requirements for Internet quality of service.In: Turner T, Szwillus G, eds. Proc. of the Conf. on Human Factors in Computing Systems. New York: ACM Press. 2000. 297−304.。
专业英文字汇
Trigonometric function
三角函数
Inverse trigonometric function
反三角函数
Naturalexponentialfunction
自然指数函数
Naturallogarithmicfunction
自然对数函数
物理实习30
dynaห้องสมุดไป่ตู้ics
动力学
kinematics
扇区
security
安全性
starvation
饥饿
statement
叙述
subalgorithm
子算法
subprogram
子程序
subroutine
子程序
syntax
语法
Unicode
万国码
variable
变数
whitespace
空白
frame
韧体
message
讯息
RAM
随机存取内存
ethernet
资料列
information
信息
router
路由器
network
网络
internet
因特网
recursive
递回
broadcast
广播
modem
调制解调器
register
缓存器
header
标头
compile
编译
compiler
编译器
tree
树
robot
机器人
monitor
屏幕
bandwidth
频宽
hacker
IBM Rational统一流程
专业英文字汇
byte
字节
carry
进位
client
客户端
cohesion
凝聚
collision
碰撞
compiler
编译器
compression
压缩
constant
常数
couping
耦合
database
数据库
datagram
资料框
deadlock
死结
declaration
宣告
decoding
解碼
decryption
Shift Operator
移位运算子
模数
most significant bit‚MSB
最大位
multiplexers
多任务器
one’s complement
l的补码
oscillate
振荡
output enable
输出致能
parity bit
同位
priority
优先权
product
积
quad
四个
reduction
简化
redundant
以太网络
subtree
子树
nonpreemptive
不可抢先排班
Method
方法
matching
比对
host
主机
server
主机服务器
Pascaline
加法器
encryption
加密
preemptive
可抢先排班
reliability
可靠性
portability
可移植性
readability
可读性
肖特太阳能公司之太阳能技术和产品的发展[1]
![肖特太阳能公司之太阳能技术和产品的发展[1]](https://img.taocdn.com/s3/m/aa81fa0b79563c1ec5da713e.png)
Al / Ni
Advantages （优势）:
1958 1980 1983 2002
applications (AEG)太阳能电池片应用于空间领域 Development work on thin film solar cells at MBB (a-Si) MBB对于薄膜电池片有进展 First shipments of glass tubes for solar thermal power plants (SCHOTT Rohrglas GmbH)
开始集热器系列产品规模化生产
2004 2005 2006 2007
JV Wacker SCHOTT Solar founded 成立 JV Wacker SCHOTT Solar 公司 Thin film production started in Jena 在 Jena生产薄膜组件 New sites set up in Spain and in the USA; existing sites expanded
大型系统 / 太阳能发电厂
© SCHOTT Solar AG
6
SCHOTT Solar AG
SNEC 3rd Solar (2009)at a glance SCHOTT
ASI® Products ASI 薄膜产品 Thin Film Amorphous ASI®
Framed modules 有框组件 • On-/Off-Grid 并网/离网 • 12 V applications 12V 应用
2
SCHOTT Solar AG
1 SNEC SCHOTT 3rd Solar (2009) Solar a glance SCHOTT atat a glance
fcbga封装流程介绍
fcbga封装流程介绍英文回答：Flip Chip Ball Grid Array (FCBGA) Packaging Process.The flip chip ball grid array (FCBGA) packaging process is a type of surface-mount technology (SMT) that is used to package integrated circuits (ICs). In the FCBGA process,the IC is flipped upside down and mounted on a substratewith solder balls connecting the IC to the substrate.The FCBGA packaging process begins with the fabrication of the IC. The IC is typically fabricated on a silicon wafer, and the individual ICs are then diced from the wafer. The ICs are then cleaned and prepared for assembly.The next step in the FCBGA packaging process is toapply solder balls to the IC. The solder balls aretypically made of a lead-tin alloy, and they are applied to the IC using a stencil printing process. The solder ballsare then reflowed, which melts the solder and forms a connection between the IC and the substrate.The IC is then placed on the substrate, and the solder balls are reflowed again. This step forms a permanent connection between the IC and the substrate.The final step in the FCBGA packaging process is to encapsulate the IC. The encapsulation material is typically a plastic or ceramic compound, and it is applied to the IC using a molding process. The encapsulation material protects the IC from the environment and providesstructural support.中文回答：倒装芯片球栅阵列（FCBGA）封装流程。
外文翻译林玉强
4产品全生命周期中MBD要求在本节中，与数据内容，数据访问、可视化和数据保留有关的一些问题和关注用具体要求与标准进行说明。
4.1.数据内容要求为了理解MBD数据集的内容应该是什么，我们必须先看看整个产品生命周期工程图纸所含有的信息。
根据ASME Y14.1-2005和ISO10209-1:1992，我们所定义的工程图纸4：按照国际标准和通用规模明确的形式，在二维平面的信息载体上通过静态图形和图案表现（或两者的组合），携带、控制和维护一个项目的产品定义的现场工程文件，为了满足物理和功能的要求以及多学科的约束，项目的产品定义是作为一套在特定平衡状态下共存的产品属性、功能和特性来定义的。
4术语工程制图可以视为广义词的一个子集，被称为“技术制图”，并将其应用到任何图纸来表达技术的想法。
然而，在本文中，我们使用术语工程制图而不是技术制图，因为实施MBD是目前针对工程的应用。
作为文件，工程图纸具有基本属性。
根据Adamczewski，一个文件是载体、语义构造和上下文说明之间的结合。
根据这一定义，我们可以识别一个文件的两个基本属性：介质（物理支持）和信息（内容）。
根据Adamczewski，这两个属性中的任何一个都不能够定义文件。
文件中信息在语义上构造和描述的方式看作是文件格式。
在这个意义上工程图纸是一种文件，它有一个介质（纸质或电子的支持），信息（图纸的内容）和文件格式（图纸内的安排和组织信息）。
这三个文件方面，介质，内容和格式，在下节进行细节讨论。
4.1.1.内容方面传统上，工程图纸表示信息单元，满足客户所需要的多学科下所有的产品定义，因此，所需的信息要涵盖广泛的题材。
基于美国机械工程师协会的y14.1 - 2005[30]和y14.41-2003[11]标准，我们建议把图纸元素分成三个不同的类别：(a)产品定义的核心元素，(b)产品定义的外围要素和(c)产品定义的管理要素。
所有这些元素是可见的，无需任何手动或外部操作。
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
ＶＯ１．１６ ＮＯ．２ ＣＡＤＤＭ Ｄｅｃｅｍｂｅｒ ２００６ Ｒｅｓｏｕｒｃｅ Ｅｎｃａｐｓｕｌａｔｉｏｎ ｉｎ Ｍａｎｕｆａｃｔｕｒｉｎｇ Ｇｒｉｄ Ｅｎｖｉｒｏｎｍｅｎｔ ＬＩＹｕ．ＣＨＥＮＹｏｕ－ｐｉｎｇ，ＺＨＯＵ Ｚｕ‘ｄｅ （Ｓｃｈｏｏｌ ｏｆ Ｍｅｃｈａｎｉｃａｌ Ｅｎｇｉｎｅｅｒｉｎｇ，Ｈｕａｚｈｏｎｇ Ｕｎｉｖｅｒｓｉｔｙ ｏｆ Ｓｃｉｅｎｃｅ ａｎｄ Ｔｅｃｈｎｏｌｏｇｙ Ｗｕｈａｎ ４３００７４。Ｃｈｉｎａ） Ａｂｓｔｒａｃｔ：Ａ ｆｅａｓｉｂｌｅ ｈｉｅｒａｒｃｈｉｃａｌ ｓｔｒｕｃｔｕｒｅ ｏｆ ＭＧ ｉｓ ｉｎｔｒｏｄｕｃｅｄ ａｎｄ ａｎ ＸＭＬ ｆｏｒｍａｔ ｔｏ ｄｅｓｃｒｉｂｅ ｍａｎｕｆａｃｔｕｒｉｎｇ ｒｅｓｏｕｒｃｅ，ａｎ ｅｍｂｅｄｄｅｄ ＮＣ ｓｙｓｔｅｍ ｉｓ ｉｌｌｕｓｔｒａｔｅｄ．Ａｎ ｅｘａｍｐｌｅ ｏｆ ｅｎｃａｐｓｕｌａｔｅｄ ｒｅｓｏｕｒｃｅ ｓｈａｒｉｎｇ ｉｓ ｇｉｖｅｎ ｔｏ ｄｅｍｏｎｓｔｒａｔｅ ｔｈｅ ｐｒｏｃｅｓｓｅｓ ｉｎｖｏｌｖｅｄ ｉｎ ｒｅａｌｉｚｉｎｇ ａ ｕｓｅｒ，Ｓ ｒｅｑｕｅｓｔ ｆｏｒ ａ ＣＡＤ ｒｅｓｏｕｒｃｅ ｓｅｒｖｉｃｅ ｔｈｒｏｕｇｈ ｐａｒａｌｌｅｌ ｄｅｓｉｇｎ ａｎｄ ｃｏｍｐｕｔｉｎｇ・
Ｋｅｙ ｗｏｒｄｓ：ｍａｎｕｆａｃｔｕｒｉｎｇ ｇｒｉｄ；ｒｅｓｏｕｒｃｅ ｓｈａｒｉｎｇ；ＯＧＳＡ；ｒｅｓｏｕｒｃｅ ｍｏｄｅｌ，ｄｅｓｃｒｉｐｔｉｏｎ ａｎｄ ｅｎｃａｐｓｕｌａｔｉｏｎ ＣＬＣ ｎｕｍｂｅｒ：ＴＨ １６５
Ａｒｔｉｃｌｅ ＩＤ：１００３－４９５１（２００６）０２・－００２３・－０７
１ ｌｎｔｒｏｄｕｃｔｉｏｎ Ａｓ ａ ｎｅｗ ａｐｐｒｏａｃｈ， ｄ ｔｅｃｈｎｏｌｏｇｙ ｈａｓ ｂｅｅｎ ｒａｐｉｄｌｙ ｕｓｅｄ ｉｎ ｓｃｉｅｎｔｉｆｉｃ ｃｏｍｐｕｔｉｎｇ，ｌａｒｇｅ。ｓｃａｌｅ ｄａｔａ ｍａｎａｇｅｍｅｎｔ，ａｎｄ ｃｏｌｌａｂｏｒａｔｉｖｅ ｗｏｒｋ．Ｂｕｔ ｉｎ ｔｈｅ ｆｉｅｌｄ ｏｆｍａｎｕｆａｃｔｕｒｉｎｇ，ｔｈｅ ａｐｐｌｉｃａｔｉｏｎ ｏｆ ｄ ｉｓ ｊｕｓｔ ａｔ ｔｈｅ ｂｅｇｉｎｎｉｎｇＩ“．Ｉｎｖｏｌｖｅｍｅｎｔ ｉｎ ｔｈｅ ｄｅｖｅｌｏｐｍｅｎｔ
ｏｆ ＭＧ ｃｏｕｌｄ ｂｒｉｎｇ ｉｎ ｓｉｇｎｉｆｉｃａｎｔ ａｄｖａｎｔａｇｅｓ ｂｙ ｍａｋｉｎｇ ｂｅａｅｒ ｕｓｅ ｏｆ ｅｘｉｓｔｉｎｇ ｍａｎｕｆａｃｔｕｒｉｎｇ ｃａｐａｃｉｔｙ ａｎｄ ａｂｉｌｉｔｉｅｓ．Ｉｆ ｔｅｃｈｎｉｃａｌｌｙ ａｎｄ ｅｃｏｎｏｍｉｃａｌｌｙ ｆｅａｓｉｂｌｅ，ａ ＭＧ ｐｒｏｍｉｓｅｓ ｔｏ［２】：
（１）Ｍａｘｉｍｉｚｅ ｔｈｅ ｕｓｅ ｏｆ ｅｘｉｓｔｉｎｇ ｍａｎｕｆａｃｔｕｒｉｎｇ ｓｕｐｐｌｙ ｃｈａｉｎ ｒｅｓｏｕｒｃｅｓ； （２）Ｅｎａｂｌｅ ｈｉｇｈｅｒ ｌｅｖｅｌｓ ｏｆ ｃｕｓｔｏｍｅｒ ｒｅｓｐｏｎｓｅ ａｎｄ ｃｕｓｔｏｍｉｚａｔｉｏｎ； （３）Ｐｒｏｍｏｔｅ ｈｉｇｈｅｒ ｅｆｉｃｉｅｎｃｙ ｉｎ ｅｘｉｓｔｉｎｇ ｓｕｐｐｌｙ ｎｅｔｗｏｒｋｓ；
Ｄｏｃｕｍｅｎｔ ｃｏｄｅ：Ａ （４）Ｃｏｐｅ ｗｉｔｈ ｍｏｒｅ ｃｏｍｐｌｅｘ ｓｙｓｔｅｍｓ； （５）Ｅｎａｂｌｅ ｏｒｇａｎｉｚａｔｉｏｎｓ ｔｏ ｓｕｒｖｉｖｅ ｉｎ ｔｈｅ ｆａｃｅ ｏｆ ｕｎｐｒｅｄｉｃｔａｂｌｅ ｃｕｓｔｏｍｅｒ ｄｅｍａｎｄ； （６）Ｅｘｐｌｏｉｔ ｎｅｗ ｉｄｅａｓ ｒａｐｉｄｌｙ． ＭＧ ｉｓ ｔｈｅ ｄｅｖｅｌｏｐｉｎｇ ｐｌａｔｆｏｒｍ ａｎｄ ｓｕｐｐｏｒｔｉｎｇ ｅｎｖｉｒｏｎｍｅｎｔ ｆｏｒ ｍｏｄｅｒｎ ｉｎｔｅｇｒａｔｅｄ ｍａｎｕｆａｃｔｕｒｉｎｇ ｓｙｓｔｅｍ，ｗｈｉｃｈ ｃｏｍｂｉｎｅｓ ｒｅｌａｔｅｄ ｄｅｓｉｇｎ ａｎｄ ｍａｎｕｆａｃｔｕｒｉｎｇ ｒｅｓｏｕｒｃｅｓ，ｐｒｏｃｅｓｓｅｓ ａｎｄ ｋｎｏｗｌｅｄｇｅａｂｌｅ ｉｎｆｏｒｍａｔｉｏｎ ｉｎ ｔｈｅ ｗｈｏｌｅ ｌｉｆｅ ｏｆ ｐｒｏｄｕｃｔｉｏｎ．Ｉｔ ｐｒｏｖｉｄｅｓ ｍａｎｕｆａｃｔｕｒｉｎｇ ｅｎｔｅｒｐｒｉｓｅｓ ｗｉｔｈ ｍａｎｕｆａｃｔｕｒｉｎｇ ｒｅｓｏｕｒｃｅｓ ａｎｄ ｓｅｒｖｉｃｅｓ ｉｎ ｔｈｅ ｓａｍｅ ｗａｙ ａｓ ｔｈｅ Ｉｎｔｅｒｎｅｔ ｐｒｏｖｉｄｉｎｇ ｉｎｆｏｒｍａｔｉｏｎ ｓｅｒｖｉｃｅｓ．Ｉｔ ａｌｓｏ ｍａｋｅｓ ｅｎｔｅｒｐｒｉｓｅｓ ａｃｑｕｉｒｅ ａｌｌ ｔｈｅ ｓｅｒｖｉｃｅｓ ｒｅｌａｔｅｄ ｔｏ ｍａｎｕｆａｃｔｕｒｉｎｇ ｃｏｎｖｅｎｉｅｎｔｌｙ，ａｎｄ ｕｓｅ ａｌｌ ｔｈｅ ｒｅｓｏｕｒｃｅｓ ｅｎｃａｐｓｕｌａｔｅｄ ｉｎ ＭＧ ａｓ ｅａｓｉｌｙ ａｓ ｕｓｉｎｇ ｔｈｅ ｌｏｃａｌ ｏｎｅｓ．Ｔｈｅ ｕｌｔｉｍａｔｅ ｏｂｊｅｃｔｉｖｅ ｏｆ ＭＧ ｉｓ ｔｏ ｒｅａｌｉｚｅ ｒｅｓｏｕｒｃｅ ｓｈａｒｉｎｇ，ｃｏｌｌａｂｏｒａｔｉｖｅ ｄｅｓｉｇｎ ａｎｄ
Ｒｅｅｅｉｖｅｄ ｄａｔｅ：Ｏｃｔｏｂｅｒ ２０，２００６ Ｆｏｕｎｄａｔｉｏｎ ｉｔｅｍ：Ｓｕｐｐｏｒｔｅｄ ｂｙ ｔｈｅ Ｎａｔｕｒａｌ Ｓｃｉｅｎｃｅ Ｆｏｕｎｄａｔｉｏｎ ｏｆ Ｃｈｉｎａ（５００３３５０２０） Ｂｉｏｇｒａｐｈｙ：ＬＩ Ｙｕ（１９７４－），ｍａｌｅ，ｂｏｒｎ ｉｎ Ｘｉｎｇｎｉｎｇ ｃｏｕｎｔｙ，Ｇｕａｎｇｄｏｎｇ ｐｒｏｖｉｎｃｅ，Ｐｈ．Ｄ ｃａｎｄｉｄａｔｅ，ｍａｉｎ ｒｅｓｅａｒｃｈ ｆｉｅｌｄｓ ａｒｅ ｉｎｔｅｌｌｉｇｅｎｔ ｍａｎｕｆａｃｔｕｒｉｎｇ， ｖｉｒｔｕａｌ ｒｅａｌｉｔｙ ａｎｄ ｄｉｇｉｔａｌ ｔｅｘｔｉｌｅ ｔｅｃｈｎｏｌｏｇｙ．１ｅｅｗｏｏ＠ｗｕｓｅ．ｅｄｕ．ｃｎ
维普资讯 http://www.cqvip.com ＬＩ Ｙ ，ＣＨＥＮ Ｙｏｕ－Ｐｉｎｇ，Ｚｈｏｕ Ｚｕ－ｄｅ ｍａｎｕｆａｃｔｕｒｉｎｇ，ｔｏ ａｃｈｉｅｖｅ ｐｅｒｆｏｒｍａｎｃｅ ｏｆ ｌｏｗｅｒ ｃｏｓｔｓ，ｈｉｇｈ ｅｆｆｉｃｉｅｎｃｙ ｏｆ ｒｅｓｏｕｒｃｅ ｕｔｉｌｉｚａｔｉｏｎ ａｎｄ ｆａｓｔ ｒｅａｃｔｉｏｎ ｔｏ ｍａｒｋｅｔｓ，ＳＯ ｒｅｓｏｕｒｃｅ ｓｈａｒｉｎｇ ｉｓ ｔｈｅ ｋｅｙ ｆ．ａｃｔＯｒ ｏｆ ＭＧｔ ． Ａｓ ＭＧ ｉＳ ｐｒｅｓｅｎｔｅｄ ａｓ ａｎ ａｄｖａｎｃｅｄ ｓｏｌｕｔｉｏｎ ｆｏｒ
ｔｈｅ ｂｏｔｔｌｅｎｅｃｋ ｏｆ ｎｅｔｗｏｒｋｅｄ ｍａｎｕｆａｃｔｕｒｉｎｇ，ｔｈｅ ｒｅｓｅａｒｃｈ ｏｎ ｔｈｅ ｒｅｓｏｕｒｃｅ ｅｎｃａｐｓｕｌａｔｉｏｎ ｉｎ ＭＧ ｅｎｖｉｒｏｎｍｅｎｔ ｗｉｌｌ ｂｕｉｌｄ ｕｐ ｓｏｌｉｄ ｔｈｅｏｒｅｔｉｃａｌ ａｎｄ ｔｅｃｈｎｏｌｏｇｉｃａｌ ｆｕｎｄａｍｅｎｔａｌｓ ｔｏ ｒｅａｌｉｚｅ ａ ｇｒｅａｔ ｓｔｒｉｄｅ ｉｎ ｍａｎｕｆａｃｔｕｒｉｎｇ．Ｔｈｉｓ ｐａｐｅｒ ｍａｉｎｌｙ ｆｏｃｕｓｅｓ ｏｎ ｒｅｓｏｕｒｃｅ ｒｅｇｉｓｔｒａｔｉｏｎ，ｅｘｐｒｅｓｓｉｏｎ ａｎｄ ｅｎｃａｐｓｕｌａｔｉｏｎ ｉｎ ＭＧ ｅｎｖｉｒｏｎｍｅｎｔ．Ｆｉｒｓｔ。ｔｈｅ ａｒｃｈｉｔｅｃｔｕｒｅ ｏｆ ｔｈｅ ｃｏｌｌａｂｏｒａｔｉｖｅ ｄｅｓｉｇｎ ｐｌａｔｆｏｒｍ ｂａｓｅｄ ｏｎ ＭＧ ｉＳ ｉｎｔｒｏｄｕｃｅｄ；ｔｈｅｎ ｍａｎｕｆａｃｔｕｒｉｎｇ ｒｅｓｏｕｒｃｅ ｍｏｄｅｌ， ｒｅｓｏｕｒｃｅ ｄｅｓｃｒｉｐｔｉｏｎ ａｎｄ ｅｎｃａｐｓｕｌａｔｉｏｎ ｂａｓｅｄ ｏｎ ＯＧＳＡ ｉｓ ｓｔｕｄｉｅｄ；ｆｉｎａｌｌｙ，ａｎ ｅｘａｍｐｌｅ ｏｆ ｅｎｃａｐｓｕｌａｔｅｄ ｒｅｓｏｕｒｃｅ ｓｈａｒｉｎｇ ｉｓ ｐｒｏｖｉｄｅｄ．
２ ＭＧ Ｐｌａｔｆｏｒｍ Ａｒｅｈｉｔｅｅｔｕｒｅ ＭＧ ｉＳ ａ ｍａｎｕｆａｃｔｕｒｉｎｇ－ｏｒｉｅｎｔｅｄ ｖｉｒｔｕａｌ ｎｅｔｗｏｒｋ ｐｌａｔｆｏｒｍ ｏｎ ｔｈｅ ｂａｓｉｓ ｏｆＩｎｔｅｒｎ瓯ｇｒｉｄ，ａｎｄ ｏｔｈｅｒ ｒｅｌａｔｅｄ ｔｅｃｈｎｏｌｏｇｉｅｓ，ａｎｄ ｉｔ ｓｈｉｅｌｄｓ ｔｈｅ ｈｅｔｅｒｏｇｅｎｅｏｕｓｎｅｓｓ ａｎｄ ｔｈｅ ｒｅｇｉｏｎａｌ ｄｉｓｔｒｉｂｕｔｉｏｎ ｏｆ ｒｅｓｏｕｒｃｅｓ ｂｙ ｗａｙ ｏｆ ｅｎｃａｐｓｕｌａｔｉｎｇ ａｎｄ ｉｎｔｅｇｒａｔｉｎｇ ｔｈｅ ｄｅｓｉｇｎ，ｍａｎｕｆａｃｔｕｒｉｎｇ，ｍａｎａｇｅｍｅｎｔ，ｉｎｆｏｒｍａｔｉｏｎ， ｔｅｃｈｎｏｌｏｇｙ，ｉｎｔｅｌｌｉｇｅｎｃｅ ａｎｄ ｓｏｆｔｗａｒｅ ｒｅｓｏｕｒｃｅｓ ｄｉｓｔｒｉｂｕｔｅｄ ｉｎ ｄｉｆｅｒｅｎｔ ｅｎｔｅｒｐｒｉｓｅｓ ａｎｄ ｇｒｏｕｐｓ［４１．Ａｓ
ＭＧ ｉｓ ｂｕｉｌｔ ｏｎ ｔｈｅ ｂａｓｉｓ ｏｆ ｔｏｄａｙ’Ｓ Ｉｎｔｅｒｎｅｔ，ｉｔ ｉｓ ｃｏｎｓｔｒｕｃｔｅｄ ｐｈｙｓｉｃａｌｌｙ ｏｎ ｔｈｅ ｇｒｏｕｎｄ ｏｆ ｔｒａｄｉｔｉｏｎａｌ ＴＣＰ／ＩＰ ｉｎｆｒａｓｔｒｕｃｔｕｒｅｓ ｗｉｔｈ ｖａｒｉｏｕｓ ｋｉｎｄｓ ｏｆ ｒｅｓｏｕｒｃｅｓ ａｓ ｉｔｓ ｎｏｄｅｓ。ａｎｄ ｉｔｓ ａｒｃｈｉｔｅｃｔｕｒｅ ｉＳ ｓｉｍｉｌａｒ ｔｏ ｔｌ１ａｔ ｏｆ Ｉｎｔｅｍｅｔ ｗｉｔｈ ａ ｈｉｅｒａｒｃｈｉｃａｌ ｓｔｒｕｃｔｕｒｅ ｔｏ ｃｅｒｔａｉｎ ｄｅｇｒｅｅ，ａｓ ｉｓ ｓｈｏｗｎ ｉｎ Ｆｉｇ．１．
Ａｐｐｌｉｃａｔｉｎ Ｉ Ａｖｉ￣ｉｏｎ ．Ａｕｔｏｍｏｂｉｌｅ Ｉ Ｓｈｉｐｐｉｎｇ Ｉ Ｔｅｘｔｉｌｅ Ｍｅｄｉｃｉｎｅ １
ｌ ＭＧ Ｐ。ｒｔａｌ ｌ
Ｕｓｅｒ Ｌｅｖｅｌ Ｍｉｄｄｌｅｗａｒｅ ｌ ＰｅｒｆｏｒｍａｎｃｅＥｖａｌｕａｔｉｏｎＴｏｏｌ ｌ Ｗｅｂ Ｔｏｏｌｓ Ｉ ＣＳＣＷ ｌ Ａｃｃｏｕｎｔ Ｉ …
ＣｏｒｅＭｉｄｄｌｅｗａｒｅ Ｉｎｆｏｒｍａｔｉｏｎ Ｒｅｓｏｕｒｃｅ Ｄａｔａ Ｓｅｃｕｒｉｔｙ ＷＯｒｌ（ｆｌＯＷ Ｍａｎａｇｅｍｅｎｔ ．Ｍａｎａｇｅｍｅｎｔ Ｍａｎａｇｅｍｅｎｔ Ｍａｎａｇｅｍｅｎｔ Ｍａｎａｇｅｍｅｎｔ
ｌ ＧＩ。ｂｕｓ Ｓｅｃｕｒｉｔｙ Ｉｎｔｅｒｆａｃｅ（ＧＳＩ） ｌ
Ｆａｂｒｉｃ
ＭＧＩｎｔｅｒｆａｃｅ Ｍ＿Ｇ Ｉｎｔｅｒｆａｃｅ ＭＧ Ｉｎｔｅｒｆａｃｅ Ｍａｎａｇｅｍｅｎｔ ａｎｄ Ｍａｎａｇｅｍｅｎｔ ａｎｄ Ｍａｎａｇｅｍｅｎｔ ａｎｄ Ｍｏｎｉｔｏｒｉｎ￣Ｓｖｓｔｅｍ Ｍｏｎｉｔｏｒｉｎ￣Ｓｙｓｔｅｍ Ｍｏｎｉｔｏｒｉｎ￣Ｓｙｓｔｅｍ ＭａｎｕｆａｃｔｕｒｉｎｇＲｅｓｏｕｒｃｅ ＭａｎｕｆａｃｔｕｒｉｎｇＲｅｓｏｕｒｃｅ ＭａｎｕｆａｃｔｕｒｉｎｇＲｅｓｏｕｒｃｅ